Simulation Study on Centralised vs Decentralised Die Bank

Approaches on Semiconductor Supply Chains

By

Kartikeya Bhardwaj
Nandish Gedia
Yungching Fan

The thesis is submitted to University College Dublin in part fulfilment of the requirements for the
degree of Master of Engineering Science in Engineering Management

School of Mechanical and Materials Engineering

Supervisor: Dr. Vincent Hargaden

27/08/2023
 Copyright Declaration
The thesis is the copyright of the authors' original research. It has been composed by the authors
and has not been previously submitted for examination which has led to the ward of a degree.

The copyright of this thesis belongs to the authors. Due acknowledgment must always be made
of the use of any of the material contained in, or derived from, this thesis.

II
Acknowledgements

We would like to express our heartfelt gratitude to all those who have contributed to the
completion of this thesis. Their support, guidance, and encouragement have been invaluable
throughout this journey.

We are deeply indebted to our supervisor, Dr. Vincent Hargaden, whose insightful feedback,
and unwavering guidance steered us in the right direction. Your expertise and dedication have
been instrumental in shaping this research.

We are also thankful to Dr. Pezhman Ghadimi for providing such a great opportunity of
working with one of the market leading firms in semiconductor industry.

We are also grateful to Mr. Liam Faulkner and Mr. Oliver Fernandez for their valuable inputs
and suggestions that greatly enriched the quality of this thesis. Their diverse perspectives added
depth to our understanding of the subject matter.

Our gratitude also goes to our family for their unwavering support and understanding,
especially during the challenging phases of this research. Your belief in us kept us going.

Lastly, we acknowledge the contributions of all the participants and resources that enabled this
research to take shape.

III
Executive Summary
The project's objective is developing a decision support tool tailored to semiconductor
manufacturing logistic planning. This tool aids in evaluating and choosing between centralized
and decentralized die bank strategies. It encompasses a thorough literature review of
semiconductor supply chains, risks, and influencing factors. Sections 1 and 3 clarify the
project's purpose and scope. Developed using Agile project management techniques, the
schedule shown in Section 3.2 outlines key milestones. In Section 4.1, the initial tool iteration
is addressed. It covers a single die bank, multiple assembly vendors, and chip types, with a plan
for future refinements. Section 4.4 advances to the next iteration, introducing a model
accommodating diverse products shipped concurrently, enhancing analytical capabilities. After
the complete development of the model, all the results were analysed, followed by which
sensitivity analysis is performed where the initial analysis compares shipping, storage, and
delay costs between the two approaches. It then evaluates delay and rerouting probabilities in
the decentralized method. Lastly, it outlines the factors influencing the decision-making
process.

In conclusion, this project targets an adaptive decision support tool for semiconductor
manufacturing. It navigates centralization vs. decentralization die bank options, grounded in
supply chain insights. The Agile project management methodology used for developing the
model ensures flexibility. The timeline and iterative enhancements reflect precision and
industry relevance.

IV
Table of Contents
Copyright Declaration ............................................................................................................................ II
Acknowledgements ............................................................................................................................... III
Executive Summary .............................................................................................................................. IV
Table of Contents ............................................................................................................................... V
List of Figures ....................................................................................................................................... VI
1. Introduction ..................................................................................................................................... 1
1.1 Background ................................................................................................................................... 1
1.2 Problem Statement and Project Description ................................................................................. 2
2. Literature Review............................................................................................................................ 4
2.1 Semiconductor Supply Chain........................................................................................................ 4
2.2 Centralised/Decentralised Approaches in Semiconductor Supply Chain ..................................... 5
2.4 Supply Chain Risks ....................................................................................................................... 6
2.5 Risk Mitigation Strategies............................................................................................................. 7
2.6 Project Management Methodology ............................................................................................... 8
3. Methodology ................................................................................................................................... 9
3.1 Project Scope Management ........................................................................................................... 9
3.1.1 Collect Requirements ............................................................................................................. 9
3.1.2 Define Scope .......................................................................................................................... 9
3.1.3 Create Work Breakdown Structure ...................................................................................... 10
3.2 Project Schedule Management .................................................................................................... 11
3.3 The Optimisation Model ............................................................................................................. 12
4. Implementation ............................................................................................................................. 22
4.1 Iteration 1: Using Microsoft Excel add-in program, Solver ....................................................... 22
4.2 Iteration 2.1: Applying Mathematical Formulation .................................................................... 23
4.3 Iteration 2.2: Considering delay and rerouting cost. ................................................................... 25
4.4 Iteration 3: Considering product mix and 5-year horizon ........................................................... 26
5. Mathematical Model ......................................................................................................................... 28
6. Results and Discussion ................................................................................................................. 32
6.1 Results ......................................................................................................................................... 32
6.2 Sensitivity Analysis .................................................................................................................... 34
7. Recommendations ......................................................................................................................... 42
8. Conclusion and Future Work ........................................................................................................ 44
References ............................................................................................................................................. 47

V
List of Figures
Figure 1 Scenarios of the supply chain ...................................................................................... 3
Figure 2 Work breakdown structure for the project................................................................. 11
Figure 3 Project Gantt chart ..................................................................................................... 12
Figure 4 Process of semiconductor supply chain (Powell, Nize, & Ji, 2023).......................... 13
Figure 5 User input - Assembly site capability list (Iteration 1) .............................................. 22
Figure 6 User input - Shipping costs (Iteration 1) ................................................................... 22
Figure 7 User input - Storage and lead time costs (Iteration 1) ............................................... 23
Figure 8 Sample result table (Iteration 2.1) ............................................................................. 23
Figure 9 Modified result table (Iteration 2.1) .......................................................................... 24
Figure 10 Result table (Iteration 2.1) ....................................................................................... 24
Figure 11 User input - Reroute probability at assembly sites (Iteration 2.2) ........................... 25
Figure 12 User input - Delay probability at assembly sites (Iteration 2.2) .............................. 25
Figure 13 User input - Delay probability at sorting sites (Iteration 2.2) .................................. 25
Figure 14 Input factors (Iteration 2.2)...................................................................................... 26
Figure 15 User input - Multiple product quantity for one order (Iteration 3) .......................... 26
Figure 16 User input - Increase in storage cost on YoY basis ................................................. 27
Figure 17 User input - Increase in shipping cost on YoY basis............................................... 27
Figure 18 Input factors ............................................................................................................. 28
Figure 19 Microsoft Excel add-in program, ‘Solver’ .............................................................. 32
Figure 20 Total cost representation.......................................................................................... 33
Figure 21 Multiple product types with variable quantities ...................................................... 34
Figure 22 Centralised approach - Total cost from every sorting to every assembly site ......... 34
Figure 23 Decentralised approach - Total cost from every sorting to every assembly site ..... 34
Figure 24 Best approach from every sorting to every assembly site ....................................... 34
Figure 25 Impact of shipping cost, storage cost, and delay cost (30 days).............................. 37
Figure 26 Impact of shipping, storage, and delay costs (1 day) .............................................. 38
Figure 27 Impact of shipping, storage, and delay costs (5 days) ............................................. 39
Figure 28 Impact of delay and reroute probabilities (30 days, Decentralised, penalty lower
than reroute shipping cost) ....................................................................................................... 40
Figure 29 Impact of delay and reroute probabilities (30 days, Decentralised, penalty higher
than reroute shipping cost) ....................................................................................................... 40

VI
 1. Introduction

1.1 Background

Semiconductors, also known as integrated circuits (ICs) or microchips, play a significant role
in electronic devices, driving improvements in industries such as communications, computers,
healthcare, military systems, transportation, clean energy, and others. These semiconductors
are made from pure elements such as silicon or germanium, as well as compounds such as
gallium arsenide. The semiconductor manufacturing consists of three major phases; Front End,
Die Bank, Back End. Front end process consists of fabrication and sorting while the back end
consists of Assembly and Testing. Die banking is the technique of storing raw die and wafers
until they are needed for ASIC (Application Specific Integrated Circuit) manufacturing. This
approach assists producers in avoiding price rises, increased lead times, and different sorts of
supply chain interruptions. Manufacturers can more effectively manage the re-entry of their
semiconductor products into the supply net after any disruptions and improve production
process stability by keeping a supply of stored dies. and improve production process stability
by keeping a supply of stored die. The decision of where to locate the die bank plays a crucial
role in supply chain design of semiconductors, it may be located near the sorting site or
assembly site depending on the strategy used.

Analog Devices, Inc. (ADI) is a global semiconductor firm that bridges the physical and digital
worlds to enable Intelligent Edge innovations. They use a combination of analogue, digital, and
software technologies to provide solutions for digitised industries, mobility, digital healthcare,
and climate change mitigation. The company services a broad network of 125,000 clients
globally and has over $12 billion in revenue and a workforce of around 25,000 workers. The
supply chain network of ADI spreads over 8 countries in America, Europe, and Asia. It consists
of more than 50 factories of which 11 are internal. They are dedicated to societal impact,
sustainability, and innovation. They drive breakthroughs that influence industries and enhance
people's lives as a valued partner (Analog Devices, 2023).

1
1.2 Problem Statement and Project Description

The goal of this project is to ‘Create a thorough framework/model that will help the
business choose the best inventory and logistic approach for its die bank, whether
centralised or decentralised.’ The scenarios of the inventory and logistics of the project are
depicted in Figure 1, although this might not be universally applied to all manufacturers in the
same industry, it holds true to the company the team works with for addressing this specific
logistics and inventory planning.

The project's success relies on considering important factors such as the number of
manufacturing sites namely, sorting and assembly facilities. It is equally important to assess
transportation costs, waiting times, also known as storage time, storage expenses, the option to
switch assembly vendors, and the risks associated with production line downtime. The project's
core lies in developing a comprehensive model that incorporates these crucial variables. This
model will enable a thorough examination of how changes impact different scenarios.
Essentially, it is a way to simulate different situations and see how the overall cost would be,
and what approach would best meet the company’s goal. This will provide supply chain
decision-makers with a solid foundation for making informed choices.

The chosen methodology for the project allows flexibility in selecting the best tools for the task.
It is crucial to select tools that can effectively build the model and handle the necessary
calculations for various scenarios. Choosing the right tools ensures the project's success. At the
centre of attention is the model itself. It is more than just numbers; it narrates how different
variables influence the optimal approach to inventory and logistics management. Whether it is
consolidating inventory in one place or spreading it out to various locations, the model helps
supply chain decision-makers make well-informed decisions. Concluding the project involves
not only analysing the outcomes from the model, but also understanding their real-world
implications.

2
Figure 1 Scenarios of the supply chain

3
 2. Literature Review
The literature review explores the existing body of research on semiconductor supply chain,
centralisation, and decentralisation approaches in the industry, followed by the risks associated
with it and corresponding risk mitigation and contingency strategies.

2.1 Semiconductor Supply Chain

Mönch et al. (2018) address the various phases involved in semiconductor manufacture, the
complexity of the processes, and the supply chain approach adopted. The corporation may
outsource the wafer, or the entire product, or total manufacturing may be processed in-house,
and the supply chain would be developed in accordance with the organisation's strategy and
the way sub-contracting may be handled for smooth operation. Furthermore, it describes the
risks associated with the semiconductor industry's supply chain, notably owing to fast
technological development and demand unpredictability. It also emphasises how, while
globalisation has resulted in cost savings by transferring industry to new regions, the
complexity of supply-chain networks has expanded dramatically. The alternative approaches
that the organisation may take--'Make-to-Store' or 'Make-to-Order'-have an impact on the
supply chain design.

Sun et al. (2010) provides an overview of the supply chain network in the semiconductor
industry, as well as the three most generally employed strategies in the industry. The product
is manufactured and stored in the Push strategy, effectively drops the lead time for delivery to
the customer but increases pipeline inventory and results in higher inventory costs. The Pull
technique minimises inventory because manufacturing begins after the orders from customers
are received, but then the lead time increases. The Push-Pull approach is extensively employed
in the semiconductor industry, and the die bank serves as the centre point of the Push and Pull
strategy. Along with the various strategies, the article gives equations for calculating the
various costs connected with the supply of semiconductors, as well as the many aspects to be
considered and the design of experiments employing the various parameters. Based on the
elements evaluated, an appropriate approach or a combination of them might be executed.

Rupp & Ristic (2000) propose the notion of the virtual enterprise, which is a distributed
network of organisations that work in a chain to satisfy the end need of the clients. The article

4
describes the overall supply-chain network of the semiconductor industry, emphasising the
need of distributed planning and introducing the concepts of 'Fine Planning' and 'Local
Planning,' as well as outlining the upstream and downstream markets in the supply network.
Fine and local level planning might aid in mitigating the bullwhip effect, particularly when
producing application specific integrated circuits. The distributed planning transferred
accountability and optimisation to the local level, while the global entity maintained the overall
supply network performance and efficiency.

Sun & Rose (2015) evaluate the semiconductor supply-chain network in general, and
additionally examines the impact of modifications on the supply-chain. The authors split the
complicated supply network into three distinct levels, namely subsystem, component, and parts,
to facilitate and comprehensively examine the network. The primary goal of this study is to
create a viable framework for controlling complexity in the semiconductor supply chain. It
should be able to discover and model complexity as well as changes in a methodical manner.

2.2 Centralised/Decentralised Approaches in Semiconductor

Supply Chain

Centralised approach: Manufacturing and decision-making tasks are consolidated in one or

a few locations in a centralised approach to the semiconductor supply chain. Typically, this
entails having a primary manufacturing facility where most production tasks, like wafer
fabrication and sorting, are completed. The centralised method allows for more control and
coordination across the whole supply chain, as well as economies of scale. This case describes
having the die bank at the sorting location owned by the company and the chips are supplied
from the centralised die bank to multiple assembly sites. In cases of machine breakdowns or
other unexpected events, centralised systems have proven to be inflexible. Because of the
constant discrepancy between planning and production a crisis management must be
permanently in operation so that unexpected events can be overcome and do not propagate
disruptions down the supply chain (Rupp, 2000).

Compared to decentralised ones, centralised supply systems typically display higher supply
chain reactivity. Accordingly, centralised supply chains are typically more adaptable and able
to react to changes in consumer demand or market circumstances (Ewen, et al., 2017).

5
Decentralised approach: Manufacturing processes are spread out among several sites or
facilities in a decentralised semiconductor supply chain. This distribution may include local
assembly suppliers or manufacturing facilities near the target markets or clients. Every
decentralised site functions independently, basing localised choices on client requests and local
market traits. In this case this approach describes having die banks at one or more primary
external assembly sites. So, whatever production is performed at fabrication and sorting
location is directly shipped to these sites and is further shipped to distinct locations based on
the demand.

The effectiveness of a centralised control strategy in lowering system costs typically increases
when demand uncertainty is minimal. However, the advantages of decentralised control
become increasingly clear as demand uncertainty rises. The study also demonstrates that the
ideal degree of decentralisation is influenced by the cost of maintaining inventory.
Decentralised control can be helpful when holding costs are high for inventories since it enables
better inventory management and lower holding costs (Wang, Rivera, & Kempf, 2016).

2.4 Supply Chain Risks

Kleindorfer and Saad (2005) states that there are two broad categories of risk in supply chain
management in any type of industry: (i) risks arising from coordinating supply and demand,
and (ii) risks from disruptions to normal activities which tend to have a low probability of
occurrence yet a great consequence. Similarly, Fu et al. (2023) divide risk and uncertainty into
three parts: (i) a company’s internal uncertainty, (ii) internal supply chain uncertainty, and (iii)
external supply chain uncertainty. The first one includes product characteristics, manufacturing
processes, decision-making complexity and methods, and organisational behaviour problems.
The second one, internal supply chain uncertainty means the company has control of its
supplier partners, and risks can be the demand patterns, forecasts, and issues with infrastructure.
The last one, external supply chain uncertainty refers to factors outside of the supply chain that
involves broader issues and are normally not under any parties’ control, such as government
regulation, geopolitical situations, acts of war and terrorism, labour disputes, economic
disruptions, and natural disasters. As the semiconductor supply chain involves international
cooperation, factories, and suppliers around the globe work cooperatively and competitively to
satisfy customer needs and grow state-of-the-art technology. It is even more critical to plan and
implement disruption risk management as well as emergency response programs.

6
ON Semiconductor Corp. suffered significant disruption in 2011 when their critical Japanese
supplier suffered from a devastating earthquake and tsunami, but after that, the Arizona-based
chip maker planned and moved to new sources to secure their manufacturing process. By
identifying and evaluating the risks that the supply chain is exposed to, companies can mitigate
the exposure or allocate resources to protect the lucrative business (DiPietro, 2015). Intel
invested around $4.6 billion in a new innovative facility in Poland that serves as an assembly
and test site. The purpose is twofold: firstly, to achieve the EU's goal of regaining 20% of
semiconductor manufacturing capacity by 2030, and make the supply network geographically
balanced; secondly, to build a more resilient, responsive, and efficient supply chain as the
Polish plant is closer to the company’s existing sites in Germany and Ireland (Intel, 2023).

2.5 Risk Mitigation Strategies

To mitigate disruptions in the supply chain, the most common ways are building inventory,
and having redundant facilities and supply partners as a backup plan. Stockpiling inventory is
practical when products have low holding costs and most importantly, lower risk of
obsolescence. On the other hand, having redundant supply partners makes more sense when
goods have high holding costs and a high rate of obsolescence (Chopra & Sodhi, 2004). Even
when the overall costs are higher and capacity utilisation rates are deliberately lower, secondary
suppliers function as insurance for the business and make the network more agile and flexible
(Sheffi & Rice Jr., 2005).

Although the product characteristics and processes are different from semiconductor chips,
Dell employs repositioning, also known as transferring, their computer components between
facilities with increasing costs but decreasing lead time to mitigate delays and sudden
manufacturing planning problems. Dell even created a dedicated position, supply-routing
analyst, that deals with inventory-routing optimization. In the study, a sophisticated decision
support system with a mixed-integer program is utilised to solve the inventory and logistic
problems. The major requirement of the model was to show the trade-off between repositioning
transportation costs and the costs of stockout (Foreman, et al., 2010). The decisions in diversion
and repositioning are comparable with the rerouting practice that the project sponsor talked
about. However, in our case, the company in the semiconductor industry focuses more on the
compromise between repositioning transportation costs and storage costs at alternative sites.

7
In another supply chain disruption study, Tomlin (2006) set a threshold such that if the
disturbance lasts more than a certain period, a contingency rerouting plan shall be applied.
Meanwhile, the study considers two variables, percentage uptime and supplier capacities. The
former refers to supplier reliability, indicating how likely a rerouting plan would happen, which
is like the ‘reroute shipping possibility’ in our case; but the latter variable, supplier capacities,
is ignored in our project. Unlike our project, the rerouted quantity can be just part of the order
instead of the whole volume in this study as there might be some make-to-stock inventory at
the site already. Whereas our project requires the team to execute repositioning for an intact
purchase order without splitting the volume into different assembly sites. The results showed
that contingent rerouting can be a good tactic that reduces costs when suppliers have volume
flexibility that can increase processing capacity when needed.

2.6 Project Management Methodology

Gustavsson (2016) explains the Agile methodology of project management, which is widely
used for software development, and conducts a detailed literature assessment of the
implementation of the Agile idea in non-software development processes. It emphasises the
adaptability of the Agile technique to the project as opposed to the inflexible traditional method.
It goes on to describe the many sectors and companies that adopt Agile approach, as well as
the advantages it has over the old technique. The author gives many papers that illustrate the
effective implementation of Agile in various fields.

Franková et al. (2016) introduce the concept of Agile project management technique, explains
the approach in detail, and defines agility as the ability to effectively market economical,
superior goods with quick turnaround times and in different capacities that provide increased
value for customers by adapting to the needs of the user. It supports the strategy to be used in
projects with big data and contrasts the merits and downsides of utilising Agile and traditional
project management approaches in the context of big data.

8
 3. Methodology

3.1 Project Scope Management

3.1.1 Collect Requirements

Applying the Agile project management concept, at the initial stages of planning, project
requirements must be collected and assessed, and it is likely that the process would be iterative
as requirements might be clarified and altered during the project. Even though high-level
objectives were already conveyed in the kick-off meeting, the detailed deliverables and
expectations are gathered and managed at later stages using techniques such as brainstorming,
interviewing stakeholders and experts, benchmarking, and prototyping (Heldman, Mangano,
& Feddersen, 2018).

The team stays connected with the project sponsor and asks questions regarding the input data,
factors, and their impact, so that the importance of those components can be reflected in the
model. In the meantime, the team invests time in understanding available supply chain models
on various software platforms to learn more about the subject and know the logic behind this
type of solution and their performance. Lastly, we created a foundation with Excel as our
functional prototype. It is imperative and helpful for the team to discuss further with the
industry partner as the factors and initial results we put together can be understood easily and
quickly, and the industry partner can point out potential issues that the team has not considered
in the prototype.

3.1.2 Define Scope

Project scope statement is an essential part of the project as it documents the objectives and
deliverables, and more importantly constraints, assumptions, the work required to produce the
results, acceptance criteria and exclusions (Heldman, Mangano, & Feddersen, 2018). In our
case, there are constraints in both time and cost as the project only lasts for ten weeks and is
conducted without any financial support. The basic assumption is that the team takes the
baseline factors shared by the sponsor into consideration, but after the project the sponsor can
still modify the simulation model on their own to incorporate more elements to better suit their
needs. The acceptance criteria are having a model that (i) takes inputs of an order and then

9
generates an optimal recommendation for the supply chain problem, namely centralised and
decentralised approaches; (ii) showing alternative logistics with their total costs; (iii) has
flexibility for users to modify the weights of the existing risk probabilities.

Yet the exclusions of the model are (i) having full flexibility for users to change the formula
for computing the output, (ii) allowing new scenarios that have different conditions and
elements that was not revealed, discussed, and requested during the project development period
to be analysed.

3.1.3 Create Work Breakdown Structure

The Work Breakdown Structure is a method to decompose the final project output into tasks
and deliverables. The first level is the major project deliverable. The second level is the project
phases for the development and the sectors for the written report, which include literature
review, mathematical model, software platform, and the final project closure report. The lowest
level is the work package level which shows the detailed tasks and deliverables. In our case,
the team discussed and defined the topics that are most relevant and must be covered in the
literature review, followed by the mathematical model as our foundation for implementing the
model in a commercial software platform. It covers factors, data from the sponsor, detailed
formulas, framework, and result analysis. The tasks for using the software tool emphasise more
on the skills of realising the mathematical model with the software functions, and through
several iterations of building, testing, and modifying to make the model practical for users and
meet the requirements.

10
 Figure 2 Work breakdown structure for the project

3.2 Project Schedule Management

The overall tasks and activities sequence and their end dates are scheduled and monitored using
a Gantt chart as it clearly shows the sequences and relationships between different tasks and
their planned duration. It also gives the project team an idea if some parts are delayed or ahead,
therefore some actions can be planned accordingly after daily or weekly reviews. The major
milestones and checkpoints are labelled, and the tasks that require time and resources form
external, i.e., industry partner and/or academy supervisor, are highlighted in yellow. This
reminds the project team to inform the stakeholders in advance. The tasks shown on the chart
follow the work breakdown structure closely and therefore they are not listed in this text, only
that the chart shows more information regarding proofread periods, report submission dates,
and preparation for final presentation.

The bars indicate the planned starting and ending dates, as well as a percentage indicating the
overall progress and status to assist the team in time management.

11
 Figure 3 Project Gantt chart

3.3 The Optimisation Model

The team discussed with the industry partner and identified that the chief factors fall into two
categories, one is cost related aspects, and another is physical constraints. The former consists
of holding costs, transportation costs including possible rerouting costs, and lead time, also
known as waiting time in this case, that can be converted to penalty costs or additional costs
for this optimisation modelling purpose; and the latter is to do with decisions that are subject
to limits of the number of die banks and downstream partners, also called assembly sites, and
the technical capability of those sites. Some costs and constraints are not considered herein,
such as the storage capacity of a die bank, processing capacity at various downstream partners,
and manufacturing costs at each assembly site. The supply chain model aims at giving decision-
making suggestions on whether die banks at front end process sites, which is centralised, or
direct transportation, also known as decentralised, should be utilised in different circumstances
to achieve cost efficiency. The recommendations shall be based on minimising the total cost
by considering above mentioned supply network factors. The model is built based on all costs
and factors the industry partner shared, however, without the historical data and business
information, no data analysis can be conducted before modelling, and the relationships amongst
the objective function and variables are solely assumed logically by the project team.

A five-step approach is used to achieve this goal:

12
(I) Understand the issues and scenarios: The semiconductor industry is one of the most
complex industries, not only due to more than 500 processing steps involved in the
manufacturing and various products, but also the harsh environment it faces, e.g.,
the volatile electronic market and the unpredictable demand (Can Sun, 2015).
Figure 4 below shows the entire process of the industry, from research and design,
manufacturing, to assembly for specific applications (Powell, Nize, & Ji, 2023).
The centralized technique, where the die bank is at the sorting location and the order
is dispatched depending on demand, is the first of the two main situations
considered in this model. In contrast, in the second scenario—the decentralized
approach—primary assembly sites serve as die banks, and all orders are delivered
to those locations directly after the processes at sorting locations.

Figure 4 Process of semiconductor supply chain (Powell, Nize, & Ji, 2023)

(II) Identify factors and then quantify each of their impact: By thorough research major
factors such as storage costs, shipping costs, delay possibilities, rerouting
possibilities etc. on which the semiconductor supply chain is dependent were
identified. The basis of assumptions for all these factors are briefly explained below.
The three forms of costs, international delivery for both approaches, storage fees at
different areas, and delay penalties are the major factors when considering short-
term one-off orders and long-term aggregate plans. In this framework, the project
team did not have access to the company’s logistics and orders data as some of the
information is considered trade secrets, and the purpose for the model is to reflect

13
 the impact of core factors and make it comprehensible and efficient for users to
make decisions. That is to say, the structure and logic behind the model can explain
the situations and results regardless the historical data and events this industry has
been through.
However, in order to better capture the impact of those factors occurring in reality
and simulate data that is closer to the real world operation, the team assigned
countries in different continents that have presence of semiconductor manufacturing
sites (Analog Devices, 2023), (Powell, Nize, & Ji, 2023), and the corresponding air
freight shipping costs were found using the Rate Calculator from FedEx (2023), a
multinational transportation, e-commerce and business services company; and the
storage costs were calculated based on consideration of different factors i.e. wafer
carriers and racks, cleanroom facilities (Vietnam Cleanroom, 2022), humidity and
temperature control, security and access control, handling and transport and
inventory management and tracking. Also, a thorough research was made for
assuming the total storage costs for different countries.

The overall delay cost can be defined as a fixed amount of money or portion of the
contract price, but past studies do not determine the fee as it depends on other
business factors (Niemi, Hameri, Kolesnyk, & Appelqvist, 2020). The model in
question gives users flexibility to change this cost as one of the input factors.

(III) Set an objective: The objective to be achieved by the model were set, the first and
foremost being the model to provide a flexible framework that could be modified
easily as and when required. The objective function is to minimise the total costs
and determine the optimal supply chain decision for the order in question.

(IV) Build a Mathematical Model: The strong mathematical model at the centre of the
study closely connected several crucial elements of the system. This model serves
as a foundation for methodically evaluating the relationships and interdependencies
between various variables. Due to practical data constraints, informed assumptions
were included to compensate in data gaps and uncertainties. These presumptions
were based on historical patterns and domain knowledge found in literatures and
industry news, which allowed the model to replicate system behaviour. Insights,
improved analysis of interactions, and better decision-making were all made

14
 possible by assumptions. Instead of making certain predictions, the model's outputs
gave prospective scenarios that shed insights on a variety of probable outcomes.
The mathematical model, in essence, connected concepts from theory and actual
decision-making. It concluded problems by employing organised assumptions and
computations, allowing users to comprehend and address the system's complexity.
The detailed final mathematical model is depicted in Section 5. Mathematical
Model.
(V) Iterate with modification and validation: The team commenced the process by
developing a basic mathematical model that included relevant cost factors. Using
Microsoft Excel add-in program, 'Solver', the team methodically sought the best
option to minimise expenses for a certain product. This method revealed a clear path
for shipping and inventory decisions, which aligned with the aim of cost reduction
while maintaining operational needs. The team iteratively updated its model,
methodically changing variables and solution approaches in a cyclic process. These
iterations allowed for an extensive evaluation of various solutions, allowing the
team to gradually settle on viable methods. Importantly, the offered solutions were
carefully validated by project sponsors, who provided real-world insights and
validation, hence increasing the model's practical usefulness.
Since the project was iterative, the dynamic interaction between theoretical and
practical was captured, ensuring that the model was regularly adjusted to the
intricacies of the actual world. This collaborative and validated iterative loop
produced solutions that were not only theoretically optimum but also functionally
successful.

The inputs for every order that requires an optimal solution are (i) the quantities of product
types (pieces of wafer); (ii) the expected storage period, shown as waiting time in the table
(days); (iii) the shipping cost from a die bank to an assembly site, and rerouting costs that
represents fees from one assembly site to another assembly site (dollars per box), (iv) unit
storage costs at each sorting and assembly site (dollars per box), (v) delay cost for each logistic
decision (dollars per order); (vi) the possibility of delay (percentage); (vii) the possibility of
rerouting from an assembly site to another assembly site (percentage); (viii) the assembly site
capability matrix with various product types (binary).

15
Shipping Costs and Rerouting Cost: Shipping costs in supply chain relate to the charges
involved in transferring items or products from one point to another within the supply chain
network. It is an important component of overall logistics costs and may have a substantial
influence on the efficiency and profitability of a supply chain.

Rerouting cost refers to the charges spent when resources are diverted from their intended
course to an alternate route owing to interruptions or optimisation requirements, as is typical
in transportation, supply chains, and communication networks. Rerouting probability
represents the chances of product being relocated from one of the assembly sites to another. It
could be found from the historical data and considering some qualitative factors like geopolitics,
stability and some other natural hazards in the respective regions. Effective risk management
reduces interruptions and makes use of technology to make more informed decisions. This cost
belongs to shipping expenses, yet it is associated only in case of decentralised die bank
approach, and hence it is necessary to assign a separate parameter for the modelling purpose.

Shipping costs, unlike many other services, vary significantly as time goes on even when the
parcels’ density, weight, and freight classification stay the same. The main causes that influence
it are (i) fuel prices; (ii) price transparency due to the internet; (iii) new technologies; and (iv)
unpredictable disruptions (Weiland, n.d.). The fluctuating price of fuel immediately influences
the shipping rates as the fuel prices are one of the prime components of transportation costs in
air and the following ground parcel shipping, and the service providers can only pass the
expenses of fuel over to the consumers (Highways Today, 2022).

Back in 2021 and 2022, most of the well-known carriers, FedEx, DHL, UPS, to name a few,
stated 4.9% and 5.9% increase in general rates respectively (Garland, 2022). In the middle of
2022, both FedEx and UPS announced an average of 6.9% rates increase effective in 2023,
which was the highest General Rate Increase the two global carriers have taken in recent years.
Moreover, other surcharge fees such as third-party billing fees, peak surcharges, fuel, and
oversized surcharges increased as well, making the overall increase in what businesses are
paying more than 9%. Traditionally, those surcharges were only accounted for 10% to 15% of
the transportation spend, but they have gradually grown to become a substantial part to as much
as 40% in some cases (Garland, 2022). The latest rates reflect the fuel prices and unpredictable
disruptions mentioned above, as the service providers have seen declining volumes, rising
operating costs from inflation, post-covid transformation, and the impact of the Russian
invasion of Ukraine (PartnerShip, 2022), (Scheinberg, 2022). Fluctuated fuel costs, rerouting

16
efforts, and constrained carrying capacity were evident since the global pandemic, but the war
in Ukraine exacerbated the situation due to port congestion and longer transit times, forcing
some businesses to change their regular rail and marine cargo to air freight (United Nations
UNCTAD, 2022). However, in the meantime, the transport capacity also decreased because of
the sanctions imposed on Russian airlines, and following by the increasing overall flight times,
for the detour avoiding passing through Russian airspace (Röhlig Logistics GmbH & Co. KG.,
2022).

It is notable that some countries have higher shipping rates regardless of the distance of
receiving locations, and the contrariwise routes unexpectedly cost much less even though many
key elements namely, routes, package, manpower, and time remain the same. In other words,
the cost discrepancy from countries does not show the straightforward linear relation between
distance and price. The disparities come from several aspects: the frequency and efficiency of
services, income per capita, local trade processing costs and regulations, local operating costs,
and hedging against risk exposure associated with exchange rate all play a role in the costs
(Inanc & Zachariadis, 2012). One of the operating costs is empty containers handling. Caused
by global trade imbalance, repositioning and re-allocation of containers and the last mile fleet
management become challenging logistics problems in the transporting sector. Even when
horizontal cooperation strategies like exchange, alliances, and resource pooling have been in
use, the problem has persisted since the fundamental cause of trade imbalance is hard to be
eliminated (Abdelshafie, Salah, Kramberger, & Dragan, 2022), (Zhou & Lee, 2009). In the
study of shipping routes linking Europe to the rest of the world conducted by Oliveira and
Figueiredo (2014), the variations between inward and outward rates come from imbalance
supply and demand, consequently, service providers commonly manipulate the pricing strategy
and impose ancillary charges to stimulate business growth and retain their profit. Although
most of the research focuses on maritime freight, the findings are applicable to air shipping as
empty containers, ground shipping, and handling surcharges also occur in airline cargo services.

The above-mentioned causes make delivering fees vary in different routes and gradually
increase over time. Therefore, the global supply chain department must review the logistics
costs based on latest news and headwinds affecting destinations and incorporate them into the
model before planning further.

Storage Costs: In the semiconductor business, inventory costs are the costs related to storing
and maintaining semiconductor components, raw materials, work-in-progress (WIP), and

17
finished goods along the supply chain. These costs, which have a direct impact on a
semiconductor firm's financial stability and competitiveness, can be very high. Here, just the
product kept in a die bank is considered for the storage (inventory) cost. The die banks' storage
capacity was thought to be endless. Commented [Kb1]: I have tried to re-phrase this para.
Please check now

Assumed scenario: Commented [YCF2R1]: Ok!!

Commented [YCF3]: I'd like to move this sector to 3.3
- Type of Wafers: 300mm silicon wafers used in advanced semiconductor
Commented [NCG4R3]: Yes was thinking the same
manufacturing. Commented [PP5R3]: Is it okay now?
- Duration of Storage: Annual cost estimate.
- Storage Method: Cleanroom facilities with humidity and temperature control.
- Other Considerations: Security, handling, inventory management.
Estimated Costs:
i. Wafer Carriers and Racks: $50,000 (including initial purchase and maintenance)
ii. Cleanroom Facilities: $100,000 per year (rental, utilities, maintenance)
iii. Humidity and Temperature Control: $30,000 per year
iv. Security and Access Control: $20,000 per year
v. Handling and Transport: $15,000 per year
vi. Inventory Management and Tracking: $10,000 per year
Total Estimated Cost: $225,000 per year

The figures are hypothetical and provided for illustration purposes only. The actual
costs can vary significantly based on factors such as the location, facility size, scale
of operations, level of cleanliness required, and market conditions.

The cost of storing semiconductor wafers in warehouses can vary based on a range of factors.
Storing wafers is a critical part of the semiconductor manufacturing and distribution process,
and the choice of storage method depends on factors such as the type of wafers, the sensitivity
of the materials, the required environmental conditions, and the intended duration of storage.
Some considerations and potential costs associated with storing semiconductor wafers:

i. Wafer carriers and racks: Wafers are often stored in specialized carriers such as
cassettes, UV tapes, and racks that provide protection from physical damage and
contamination (China Patent No. CN202220648642U 20220323 , 2022). The cost of
these carriers and racks can vary based on their design, materials, and capacity.

18
 Additionally, the space required to store the carriers and racks in a controlled
environment would contribute to the overall cost (Goodman, 1991).
ii. Cleanroom facilities: Semiconductor wafers are highly sensitive to environmental
conditions, including temperature, humidity, and cleanliness. Storing wafers in
cleanroom facilities ensures a controlled environment, but cleanroom maintenance and
operation can be costly (Vietnam Cleanroom, 2022).
iii. Humidity and temperature control: Before packaging process, the semiconductor
products are sensitive to humidity and temperature fluctuations using vacuum packages
and Nitrogen cabinets (Lin, Zargar, Hu, Lin, & Leggett, 2022). Implementing humidity
and temperature control systems in the warehouse can increase storage costs.
iv. Security and access control: Semiconductor wafers are valuable assets, so security
measures such as surveillance, access control, and anti-theft measures may add to the
storage cost.
v. Handling and transport: Storing wafers may involve specialized handling equipment
and procedures to prevent damage during loading, unloading, and transportation within
the warehouse.
vi. Inventory management and tracking: Implementing effective inventory management
and tracking systems to monitor the location and condition of stored wafers could incur
additional costs.
Regarding whether storing wafers in warehouses is the best way or if other methods should be
considered, it depends on the specific needs and goals of the semiconductor manufacturer and
distributor. Storing wafers in controlled warehouse environments is a common practice and can
be effective when proper precautions are taken . However, for some applications, just-in-time
manufacturing or direct shipping to customer sites might be more suitable to reduce storage
costs and the risk of damage or contamination during warehousing.

Ultimately, the choice of storage method should align with the overall supply chain strategy,
quality requirements, and cost considerations of the semiconductor manufacturer. It is
advisable to work with semiconductor industry experts and logistics professionals to determine
the most suitable and cost-effective storage solution based on your specific circumstances.

Without real world data from the company, the project team tried to find relevant costs for this
parameter, yet not very easy. An accurate cost breakdown for a specific practical scenario
involving semiconductor wafer storage would require detailed information about factors such

19
as the type of wafers, the scale of storage, location, specific facility requirements, and other
logistical considerations. However, simplified example to illustrate how the costs could be
inferenced is provided below.

Additionally, they would consider factors such as economies of scale, contractual agreements,
and potential cost savings from optimizing their supply chain and storage practices.

Delay Cost: The financial effect and extra costs incurred because of delays or disturbances in
the transit or processing of materials or commodities are referred to as delay costs in a supply
chain. These lags can happen throughout the supply chain, including during production,
shipping, distribution, and consumer delivery. Increased storage costs, missed production
deadlines, fines for late deliveries, accelerated shipping charges, inventory carrying costs, and
possibly lost sales or customer displeasure can all be considered delay costs. The goal of
effective supply chain management is to reduce delay costs through process optimisation and
timely, smooth product movement across the supply chain.

In a semiconductor product allocation study, Sun et al. (2011) consider tardiness and backorder
penalties as a part of making resources distribution decisions. The objective function is a
combination of costs incurred in generalised lot allocation that is desired to be minimized. In
their case, two variables are introduced, one is the base penalty on unfilled units in an order,
which is set at $250 per unit; another is the penalty on delivered units that misses the due date
that is assumed at $30 per unit per day. Kempf (2004) suggests a formulation including
minimizing costs and maximizing revenues that also counts penalty for not satisfying demand
as a factor that has impact on supply chain management. However, it does not reveal the range
of this cost used in the model. Similarly, Denton et al. (2006) include penalties in the objective
function for IBM’s supply and demand matching problem. The scholars further use this cost as
a priority indicator that can be arbitrarily set higher to reflect the importance of certain orders.

Delay cost in this case can be seen as a mixture of delay penalty imposed by customers in
business agreements, extra administration expenses to handle the situation, and risks of causing
cancelation of orders or even losing business in the future. But it varies from project to project
as the order quantity, product type, application, strategic plan, negotiation power, and business
relationship with a customer may result in different amounts (Niemi, Hameri, Kolesnyk, &
Appelqvist, 2020). Delay costs are considered with some probability which is assigned by the
organisation. This probability is decided by considering several factors mentioned.

20
Sorting Site: A sorting site is a critical facility in the semiconductor industry where
semiconductor components such as microchips are systematically evaluated, graded, and
grouped depending on their quality and performance. The testing is carried out in four stages
namely Electric Test and Wafer Burn In (ET & WBI), Hot and Cold Test, Repair, Final Test
and Inking (Samsung, 2018). The test findings are useful to several departments, including
production control, wafer manufacturing, product engineering, and quality control. If the yield
results are low, modifications must be done. If the yields are high or as predicted, the process
is under control and no modifications are needed. (Place, 2018).

Assembly Site: An assembly site is where separate semiconductor components are brought
together and integrated to create fully functional devices, such as microchips, in the world of
semiconductors. Wafer sawing, bonding, packaging, sealing are the steps that are performed
during assembly stage (ATN, n.d.) .Final testing is followed by assembly before transferring
the final product to the end customer.

Centralised and Decentralised Approach: Centralised Approach applies when the die bank
is situated at the sorting site, while Decentralised Approach is applied when it is placed at the
assembly site. (Refer to Section 2.2 for detailed information).

21
 4. Implementation

4.1 Iteration 1: Using Microsoft Excel add-in program, Solver

The group started with considering the different costs incurred and involved in the decision-
making process of the die bank, a sample version was made consisting of a single sorting site,
and multiple assembly sites along with different kinds of chips. The shipping costs, storage
costs, lead time and rerouting cost were taken into the consideration. Different probabilities
were assigned for rerouting based on assumptions.

Capability list a10 a11 b12 b13 b14 c15

Japan 0 1 1 1 1 1
Taiwan 1 1 1 1 0 1
Malaysia 1 0 1 0 1 0
Philippines 0 1 0 0 0 1
China 0 0 1 1 0 0
Mexico 0 1 1 0 0 1
Poland 1 0 0 1 1 0
East America 0 1 0 0 0 1
Figure 5 User input - Assembly site capability list (Iteration 1)

Different sites were considered to have varying capabilities indicated with binary numbers;
sites capable of manufacturing a certain type of semiconductor were assigned the number 1,
while sites unable were assigned the number 0.

The model was restricted to give solution for centralised and decentralised decision separately,
another constrain the model had been, it gave result only for single product type at a time.

Figure 6 User input - Shipping costs (Iteration 1)

22
 Storage cost ($/wafer/day) Lead time convert to $/order
Die bank 60 Japan 5000
Japan 150 Taiwan 2000
Taiwan 100 Malaysia 3000
Malaysia 80
Philippines 3000
Philippines 70
China 50
China 4000
Mexico 50 Mexico 4000
Poland 60 Poland 4000
East America 100 East America 2000
Figure 7 User input - Storage and lead time costs (Iteration 1)

However, after some discussions with the industry partner, the team decided to show all costs
without having Microsoft Excel add-in program ‘Solver’ function so that users can see not only
the optimal solution, but also all the different total costs with different logistic decisions, which
they can further evaluate incorporating human expert judgement.

4.2 Iteration 2.1: Applying Mathematical Formulation

The second model was developed utilising mathematical and logical operations. It concentrated
on lead time and storage cost at various sites. Then it created a sizable graphic that lists all the
various expenses for various scenarios.

The model's analysis of all the expenses involved in delivering a product where it needs to go
is crucial. This covers the price of shipping the item, the cost of keeping it in storage, and the
quantity of the item. The model also considers how competently various locations can
assemble the products.

Figure 8 Sample result table (Iteration 2.1)

The model output revealed all the costs of alternative sites for the specific kind of
semiconductor and highlighted the lowest cost site. The model initially calculates costs for all
the sites and then check the eligible site for the input order by comparing with the capability
list. The binary capability list gives output as 0 to ineligible site and 1 to eligible site thereby
representing the sites incapable with 0 in final output. Although it produced results for all types

23
of semiconductors, it complicates the process while broadening its use to a wider range of
products. Further modifications were made to the result table, which assessed the many
available alternatives and gave the result as the location with the lowest cost as well as
centralised or decentralised output.

Figure 9 Modified result table (Iteration 2.1)

To make the model more realistic, instead of only one sorting location, numerous sorting
locations were evaluated without affecting other parameters or assumptions. The new model
examined centralised versus decentralised decision costs while examining a single product at
a time and included various sorting and assembly sites. The model delivers results for each
sorting point and its associated assembly site, providing the data as a thorough cost matrix.
This matrix provides an organised overview of the costs associated with various routing
possibilities, assisting in decision-making for optimal resource allocation and routing choices
within the framework of supply chain. The result table is depicted in Figure 10.

Commented [YCF6]: Is the order correct? I feel Fig 9

Figure 10 Result table (Iteration 2.1) and 10 are the other way around.
Commented [PP7R6]: Yes they are in correct order ,
for the final version I have not added the table as kartik
needs it in his part , so instead of repeating I thought it
would be better to have that in final result part.

24
4.3 Iteration 2.2: Considering delay and rerouting cost.
Rerouting and delay costs were also included in the model to further increase accuracy. The
cost of rerouting was computed as the product of the cost and the rerouting factor. To keep the
model simple and account for the cost variation between rerouting locations, the rerouting
factor was utilised. The rerouting cost was assumed to be the same for all rerouting locations.
The rerouting weight and likelihood were combined to get the rerouting factor. For calculation
of rerouting weight, the cost of rerouting from one assembly site to all other considered
assembly site was found and normalised. For instance, the rerouting weight of India was found
by taking median of shipping cost from India to all other assembly sites, similarly median of
all assembly sites were found and sum of all medians was calculated which was used for
normalising the rerouting weight, refer to section 5.

Re-route
India China Australia Canada Thailand Netherlands Philippines Malaysia
probability
12% 13% 17% 25% 13% 15% 10% 10%

Figure 11 User input - Reroute probability at assembly sites (Iteration 2.2)

The risk of a delay was anticipated to account for the delay costs, and a predetermined fee as a
penalty per box of semiconductors was taken into consideration. The different probability of
delay at different sites is depicted in Figure 12 and Figure 13.

Delay probability India China Australia Canada Thailand Netherlands Philippines Malaysia
0.4 0.25 0.3 0.2 0.1 0.15 0.1 0.1

Figure 12 User input - Delay probability at assembly sites (Iteration 2.2)

Figure 13 User input - Delay probability at sorting sites (Iteration 2.2)

25
 Figure 14 Input factors (Iteration 2.2)

4.4 Iteration 3: Considering product mix and 5-year horizon

Following the meeting with the project sponsors, the sponsors requested a framework for
managing multiple products at once rather than just one, as well as a rough estimate of the cost
over the next five years. These requests would enable the sponsors to make decisions at various
stages of the order and, if necessary, switch from a central to a decentralised approach based
on the cost. In addition, the sponsors required a matrix that would show the approach as
centrally located or distributed throughout all qualifying locations rather than just highlighting
the least expensive route or approach.

In order to meet the demands of the sponsors, the team decided to include multiple products at
once rather than just one, shown in Figure 15. This was possible because the major costs for
each product in the supply chain were the same, and logical functions were used to deal with
the different capabilities for each site. To be clearer, if an order consists of 4 different types of
products and if Site A is not having capability to deal with one of the products from the given
input, it will show 0 cost in total cost column, and if Site B has capability to process all the
input products, it will give the total cost as an output.

Product Series a10 a11 b12 b13 b14 c15 Total Qty
Product QTY 10000 10000 10000 10000 10000 10000 1200
Y/N condition 1 1 1 1 1 1
200 200 200 200 200 200

Figure 15 User input - Multiple product quantity for one order (Iteration 3)

26
Although the semiconductor industry is fast-moving, with the high investment, high price and
long lead time characteristics, customers tend to make forecasts and aggregate plans to secure
their chip supply. To reflect the new requirement, the product type has been extended from one
at a time to multiple types, and the estimated yearly inflation rate in storage costs and shipping
costs were included in the formulas to get an estimation of the total costs in five years. Inputs
were provided to offset the 5-year estimate, including percentage increases in storage costs at
various locations and increases in transportation costs. As fuel costs is a major determinant of
shipping costs, the rise in fuel prices was equivalent at all the locations. This increase in
transportation costs was also considered when estimating the cost of rerouting. The increase in
both costs are the inputs from users, shown in Figure 16 and Figure 17.

Increase in Storage Cost % Japan Taiwan Czech Republic Germany Ireland Mexico Poland East America
Y2 0.02 0.02 0.02 0.02 0.1 0.07 0.03 0.04
Y3 0.03 0.02 0.033 0.035 0.08 0.07 0.05 0.05
Y4 0.12 0.03 0.06 0.05 0.06 0.1 0.02 0.07
Y5 0.15 0.08 0.1 0.003 0.12 0.05 0.0075 0.05

Increase in Storage Cost % India China Australia Canada Thailand Netherlands Philippines Malaysia
Y2 0.02 0.02 0.07 0.12 0 0.02 0.04 0.06
Y3 0.05 0.02 0.07 0.02 0 0.08 0.06 0.04
Y4 0.07 0.08 0.05 0.06 0.06 0.09 0.06 0.3
Y5 0.07 0.1 0.05 0.07 0.08 0.12 0.1 0.35

Figure 16 User input - Increase in storage cost on YoY basis

Figure 17 User input - Increase in shipping cost on YoY basis

The cost collected from each qualifying site was determined while taking into account
centralised and decentralised situations similar to Figure 10, and the result table was created, Commented [YCF8]: Missing the final table screenshot
here?
see Figure 28.

27
5. Mathematical Model

Figure 18 Input factors

Variables

TCc: Total Cost in Centralised Approach.

TCD: Total Cost in Decentralised Approach.

Pk: Product Quantity (Semiconductor Wafer) where, k = a, b, c, d….n.

TBQ: Total Box Quantity (Batch of 50 for each kind of product in a box).

Ζij: Total Shipping Cost from i to j.

Αij: Shipping Cost from i to j per box, where i = A, B, C, D….M, and j = 1, 2, 3, 4….z.

Γi: Total Storage Cost at site i.

28
Γj: Total Storage Cost at site j.

Βi: Storage Cost at Sorting site i per box.

Βj: Storage Cost at Assembly site j per box.

𝜎: Storage Days, the Waiting period at site i or j.

µ: Delay Cost.

DP(i): Probability of Delay at i.

DP(j): Probability of Delay at j.

CD/B: Deal Penalty per box.

Δj: Total Rerouting Cost from site j.

RW: Rerouting Cost per box.

Rfj: Rerouting Factor for location j.

Mj: Median of shipping cost from site j to all other j (From site 1 to all other site).

Ѱj: Rerouting weightage.

R(Pj): Probability of rerouting from site j.

C: Capability for processing the product, C ∈ (0,1),

Cjk = 0, if site j is not capable of processing product k.

= 1, if site j is capable of processing product k.

YSil: Percentage Increase (Year on Year Basis) increase in storage cost for location i and year
l.

YSjl: Percentage Increase (Year on Year Basis) increase in storage cost for location j and year
l.

YTl: Percentage Increase (Year on Year Basis) increase in shipping cost.

29
i: Sorting Site.

j: Assembly Site.

k: Product Series/Type.

l: Year (1, 2, 3, 4, 5).

TBQ = ∑ P ⁄50 ……. Equation 1

The first equation gives the total quantity of boxes required to be shipped, the maximum
capacity of a box is assumed to be 50 pieces, hence, the total quantity of each kind of wafer
(product) is divided by 50 and rounded up to next higher number in case of non-integer
(Entegris, n.d.).

𝑍 = 𝐴 ∗ TBQ ∗ (1 + 𝑌𝑇 ) ……. Equation 2

Equation 2 gives the total shipping cost from sorting site (i) to assembly site (j); further it also
takes into consideration the increase in shipping cost for the particular year.

𝜏 = 𝐵 ∗ 𝑇𝐵𝑄 ∗ 𝜎 ∗ (1 + 𝑌𝑆 ) …… Equation 1

𝜏 = 𝐵 ∗ 𝑇𝐵𝑄 ∗ 𝜎 ∗ (1 + 𝑌𝑆 ) ……. Equation 2

Equation 3 and 4 give the total storage cost at sorting and assembly site respectively. Storage
cost is product of quantity, storage cost per box per day, waiting period and the percentage
increase in storage cost for the year.

𝜇=𝐶 ∗ 𝐷𝑃(𝑖) ∗ 𝑇𝐵𝑄

or …… Equation 3

𝜇=𝐶 ∗ 𝐷𝑃(𝑗) ∗ 𝑇𝐵𝑄

30
Equation 5 gives the cost incurred due to delay. It considers the delay probability, delay penalty
per box and quantity of boxes for sorting and assembly sites respectively.

∆ = 𝑅 ∗ (1 + 𝑌𝑇 ) ∗ 𝑅𝑓 …… Equation 4

𝑅𝑓 = 𝜑 ∗ 𝑅(𝑃 ) …… Equation 5

𝜑 = 𝑀 ⁄∑ 𝑀 …… Equation 6

Equation 6, equation 7 and equation 8 are used to give the rerouting cost (Refer section 4.3 for Commented [NCG9]: Should I go in detail for rerouting
as that is too much and is already explained in detail in
more details). implementation section.
Commented [YCF10R9]: Yes, please do!
For Centralised Case:

𝑇𝐶 = 𝑍 + 𝜏 + 𝜇 ∗ 𝐶 jk …… Equation 7

For Decentralised Case:

𝑇𝐶 = 𝑍 + 𝜏 +𝜇+ ∆ ∗ 𝐶 jk …… Equation 8

Equation 9 and equation 10 gives total cost output for a particular order between two eligible
sites for centralised and decentralised scenarios.

𝑍 = 𝐴 ∗ 𝑇𝐵𝑄 ∗ (1 + 𝑌𝑇 ) ∗ (1 + 𝑌𝑇 ) …… Equation 9

Lasty, while calculating for next year, the equation gets modified as shown in Equation 11
which depicts the additional factors which is to be added for calculating the shipping cost for
next year.

Other elements are also modified similarly and the cost for the particular year is calculated.

The input required from the users are the quantity of product, delay cost, waiting time. The
model is kept flexible for the user to modify the cost wherever needed.

31
 6. Results and Discussion

6.1 Results
The outcomes of each iteration are discussed in this section. The supply chain of a particular
product type is the focus of the model's first and second iterations, which is also one of the
model's constraints, as was stated in the section before. The solution was provided in the form
of the best method, i.e., centralised, or decentralised, by considering all the different types of
costs and turning other aspects like delay likelihood and rerouting possibilities into
corresponding costs as well.

The team used Microsoft Excel add-in program Solver function in the initial iteration. One
sorting site and other assembly sites were considered in this model. A few restrictions were put
in the ‘Solver’ model with the goal of minimizing costs, as can be seen in the image below.

Figure 19 Microsoft Excel add-in program, ‘Solver’

A minimum cost was established after task completion and performance for a particular case,
and the best solution was then decided upon. It was a very straightforward and simple concept
with several capability limits. The model was only able to return the answer for one sorting site

32
at a time, although it was necessary to obtain the output for the optimum method from each
sorting site to each assembly site. This was the most significant constraint as per the discussions
with the industry partner. The second model was made to overcome this constraint.

The second model used mathematical and logical operations separately for centralised and
decentralised approaches to understand which would be cheaper. It also considered single
product type as the previous model. In addition, it included examining factors such as the
storage time, etc. This model addressed the accumulation of all the different costs such as the
storage cost, shipping cost, delay, and rerouting (in case of decentralised) cost, etc. which made
it very crucial.

The model included a capability matrix that denotes a site's ability to assemble specific product
types. Considering the capability of an assembly site before shipping the fabricated
semiconductor wafer was crucial due to different reasons such as the compatibility of an
assembly site with the manufacturing process, different packaging and handling requirements,
quality, and reliability of the final product etc. (Khan, Mann, & Peterson, 2021). Not only this
but the model also compared the competency of all the sorting and assembly locations with
respect to one another. The matrix of total costs from each sorting location to each assembly
location is shown in the Figure 20 below. Any assembly location with a value of zero indicates
that it is unable to assemble the product in question, in this case, Australia, Canada, Netherlands
and Philippines.

Figure 20 Total cost representation

This model's output not only listed all the potential costs but also suggested the optimum
strategy (centralized or decentralized) based on the lowest overall cost. Even while this model
was able to overcome several limitations of the prior model and produced results for all the
various types of semiconductor wafers, it still had limitations. One of the main issues was the
model's applicability because it only takes one product category into account at a time.
However, in the actual world, many items of various sizes are transported at once while
managing any supply chain. This model could get very complicated while considering a
broader range of products at once.

33
As discussed in the previous section, after a few team meetings and discussions with the
industry sponsor, a model with further advancements was developed. This model does not only
consider multiple product types with different order sizes as shown in the Figure 21 below, but
also provided the company a five-year supply chain projection plan to make more informed
decisions.

Figure 21 Multiple product types with variable quantities

Results were also provided in terms of matrices. For the centralised and decentralised approach
separately, see Figures 22 and 23, total costs were calculated, and from every single sorting
site to every single assembly site, the best approach matrix was created, as shown in Figure 24.

Figure 22 Centralised approach - Total cost from every sorting to every assembly site

Figure 23 Decentralised approach - Total cost from every sorting to every assembly site

Figure 24 Best approach from every sorting to every assembly site

6.2 Sensitivity Analysis

The sensitivity analysis is limited to a deterministic aspect as the model is built based on logical
assumptions without the historical data from the company, meaning that rather than firstly
studying data and then getting assumptions of the distributions of the parameters, the project

34
team uses presumed and arbitrary figures researched from revenant logistics websites, literature,
and industry news, to represent the parameters in the model.

In the first part of the analysis, costs of shipping, storage, and delay are compared in both
approaches, followed by the delay probability and rerouting probability in decentralised
approach, and then lastly the determining factors for the decision-making step are elaborated.

By checking one factor at a time, this method does not study interactions between inputs, but
it is simple and clear to implement and interpret as only one factor is changed and all other
factors are fixed at the base case level. The main purpose of it is to understand which factors
are the key driver of the model behaviour (Borgonovo, 2017), (Saltelli, et al., 2007). In the
local sensitivity analysis, model inputs are varied in set ranges stated in section 3.3 The
Optimisation Model, and the base case is with 1 box of products, when delay cost per box,
storage cost, and shipping cost are all set at $400, waiting time is 30 days, and the two
probabilities are at 5%. The reroute shipping costs and the weights at different sites are
calculated with the shipping service costs obtained from FedEx (2023). And to simplify the
prestation of the results, all assembly sites are assumed to be capable of all product types,
therefore the capability list shows 1 for all cells in the table.

Chart legend:

TC_Shipping changes: The total cost of the order when the shipping cost changes from the
base value, $400, to $800, and $1,200, other factors stay the same at the base levels. For the
third some case, the shipping cost changes to $600, $1,200, and $1,800.

TC_Storage changes: The total cost of the order when the storage cost changes from the base
value, $400, to $800, and $1,200, other factors stay the same at the base levels. For the third
case, the shipping cost changes to $120, $240, and $360.

TC_Delay changes: The total cost of the order when the delay cost changes from the base value,
$400, to $800, and $1,200, other factors stay the same at the base levels.

TC_Delay probability: The total cost of the order when the delay probability changes from the
base value, 5%, to 50%, other factors stay the same at the base levels.

TC_Reroute probability: The total cost of the order when the reroute probability changes from
the base value, 5%, to 50%, other factors stay the same at the base levels.

35
TC_Shipping increase %: The growth rate of the total cost of the order when the shipping cost
changes to different values.

TC_Storage increase %: The growth rate of the total cost of the order when the storage cost
changes to different values.

TC_Delay increase %: The growth rate of the total cost of the order when the delay cost changes
to different values.

TC_Delay P increase %: The growth rate of the total cost of the order when the delay
probability changes to different values.

TC_Rerote P increase %: The growth rate of the total cost of the order when the reroute
probability changes to different values.

(I) Costs of shipping, storage, and delay in both approaches

As can be seen in the mathematical model, the coefficient of the shipping cost is the
product quantity only, whereas the storage cost is multiplied by the product quantity
and waiting days, and for the delay cost, the coefficient is the product quantity and delay
probability at a front-end site. For this what-if analysis purpose, the three costs are set
at the same levels, which are $400, $800, and $1,200, to see the impact when one of
those costs increases. Since the waiting time is always greater than 1 day, and the delay
probability is always less than 100%, when the rest conditions stay the same, the effect
of changing storage costs would always be the most profound one compared to the rest
two factors. The waiting days are set as 30 days in Figure 25, but 60 and 90 days of
waiting time show the same conclusion, only with slightly different figures in the
percentage of increase that does not affect the decision-making process.

36
 Figure 25 Impact of shipping cost, storage cost, and delay cost (30 days)

This conclusion would not completely apply when extremely efficient logistic plans
happen, that is when front end deliver products immediately to the back end without
much waiting (only 1 waiting day, Figure 26), the conclusion of storage cost being the
most influential one in the total cost still holds true, but the shipping cost in this case
has the same importance as the storage cost does.

37
 Figure 26 Impact of shipping, storage, and delay costs (1 day)

If the number of waiting days is short, for instance, less than 1 week, and at the same
time the storage cost is considerably lower than shipping cost, it is possible that changes
in shipping cost play a bigger part in the total cost than the changes in storage cost does.
An example to elaborate this idea is shown in Figure 27 that the waiting time is 5 days,
and coincidentally the shipping cost, $600, is 5 times of storage cost, $120, so when all
other factors stay the same, the impact of the increase of shipping cost would give the
same effect as the increase of storage cost. The above conclusions also hold true with
decentralised approach, as the rerouting shipping cost has a less than 100% coefficient,
namely, rerouting probability, therefore, the weight of this factor is smaller than the
dominant factors, and hence the changes of rerouting shipping costs would not
contribute any significant difference in the interpretation.

38
 Figure 27 Impact of shipping, storage, and delay costs (5 days)

(II) Probability of delay and rerouting in decentralised approach

Reroute products from an assembly site to another one only occurs in decentralised
approach, however, according to the supply chain planners during some discussions, Commented [NCG11]: Is there anything backing this
claim?
the likelihood of happening is rather little. To understand if delay penalty and extra
reroute shipping cost would contribute much to the total cost, various probabilities are
assigned to examine their effect. When the delay penalty is higher than reroute shipping
cost, the changes in delay probability have more impact on the total cost, and vice versa.
Figures 28 and 29 show the two different cases that changes in reroute and delay
probabilities can lead to different outcomes when the two costs vary. As mentioned in
Section 3.3 The Optimisation Model, the delay cost would be different for every order,
hence, it is crucial for planners to identify this expense and compare it with the reroute
shipping fee so that the company understands which factor creates a greater negative
effect on the total cost. Nevertheless, both items have relatively small impact, compared
with the influences from the costs of storage. Thus, it would only be a secondary
consideration when it comes to logistics planning.

39
 Figure 28 Impact of delay and reroute probabilities (30 days, Decentralised, penalty higher
than reroute shipping cost)

Figure 29 Impact of delay and reroute probabilities (30 days, Decentralised, penalty lower
than reroute shipping cost)

(III) Determining factors

In the model, the centralised method considers the expenses for shipping from front end
to back end, storage at front end, and the delay probability at front end, whereas the
decentralised one includes the expenses for shipping from front end to back end, storage
at back end, the delay and its probability at back end, and then an extra cost of the

40
rerouting and its likelihood of happening. The same evaluating method is used here to
determine which element of the formula weighs the most, and further has a key
influence on opting for centralised or decentralised plan for certain situations.
Since waiting days are always larger than 1, it makes the storage cost the most
significant factor when supply chain planners decide which approach is more cost-
effective. That is to say, the higher the storage cost is, the better not to keep products at
the site. Hence, very often, when the storage cost at front end is higher than back end,
despite the extra rerouting cost and its possibility or chances of delaying could increase
the total cost, it is still worth it to utilize decentralised method for the order fulfilment.
And on the contrary, once the back end storage costs more than front end storage do, it
would be sensible to apply centralised method. And if both sites storage cost the same,
centralised approach would normally be favourable as it does not include the possible
rerouting cost, unless the delay probability at the front end is higher than at the back.

41
 7. Recommendations
It is advised that the company sponsor takes the following strategic initiatives to further
optimize their semiconductor supply chain considering the report's findings:

1. Integrated Capability Evaluation: They can include cross-functional capabilities in the

capability evaluation in addition to specific products. They can also think about how well their
products will work with manufacturing procedures, packaging specifications, and quality
standards. This integrated strategy guarantees that products and assembly locations are
effectively matched, minimizing production interruptions and improving overall supply chain
agility (Chien, Kuo, & Lin, 2022).

2. Integration of Real-Time Data: Real-time data streams inclusion from manufacturing

facilities, shipping partners, and market trends could be considered (Wanganoo & Shukla,
2020). The optimization model's ability to incorporate real-time data will allow the company
to react quickly to shifting circumstances like changes in transportation costs, supply chain
disruptions, or unanticipated demand spikes.

3. Scenario-Based Analysis: The constructed model could be used to conduct scenario-based

studies for a variety of possible future scenarios, some of which is shown in the sensitivity
analysis as well. These scenarios can involve shifts in geopolitics, transportation costs, effects
of environmental legislation, or production process impacts of technological improvements.
Company sponsor can proactively create plans to reduce risks and take advantage of
opportunities by evaluating alternative situations.

4. Collaborative Decision-Making: Implementing a collaborative decision-making process

that comprises cross-functional teams from several areas, such as operations, sales, finance,
and logistics. This strategy guarantees that choices consider various viewpoints and
complement overarching corporate objectives (Politi & Street Jr, 2011)

5. Dynamic Risk Management: A framework for dynamic risk management in the process of
supply chain optimization can be focused on. This framework should regularly monitor and
assess the risks related to supply chain interruptions, market volatility, geopolitical issues, and

42
transportation. A risk mitigation strategy in place ensures adaptability to unforeseen difficulties
(Ivanov & Dolgui, 2019).

6. Supply Chain Sustainability: The model can be expanded to include sustainability

indicators including carbon emissions, energy use, and trash production. The company can
reduce both the cost and the environmental effect of its supply chain. This reflects shifting
consumer demand for eco-friendly products and is consistent with the company's commitment
to sustainability (Silva, Fritz, & El-Garaihy , 2022).

In conclusion, the semiconductor supply chain optimization provides invaluable insights and
recommendations for the organisation. The multi-iteration approach, capability matrix, five-
year projection, and cost-effective approach identification collectively contribute to enhancing
company's supply chain efficiency, resilience, and strategic decision-making. By adopting the
recommendations and staying committed to continuous improvement, company can position
itself as a frontrunner in the semiconductor industry while effectively navigating its intricate
supply chain landscape.

43
 8. Conclusion and Future Work
After having some validation runs and meetings with the industry partner, the team developed
a user-friendly model with Excel presenting a matrix which shows the suitable approach
(centralised or decentralised) for various sorting and assembly sites, and in the third iteration,
further considered multiple product types of different volumes based on sales forecast over a
certain period. The deliverable is set to be a clear dashboard-like interface that simply gets
factors from users as inputs, and then aids in decision making for different orders additionally
it also has the ability for users to examine the other possible routes and their performances,
which are the total costs in this case, with auxiliary supply chain decisions. The team completed
the task mentioned above, and further did a sensitivity analysis to better understand which
factors in what scenarios have the major impact on the supply chain decisions.

Yet there are some of the factors that are excluded in the current version of framework but
could be included are mentioned further.

Rerouting Cost

Future work in this area has the potential to greatly improve the reliability and accuracy of
rerouting cost assessments. Establishing precise transport costs inside the network between
each unique site is one area for development. The method currently being used for rerouting
cost analysis centres around figuring out the median cost incurred while rerouting from one
specific site to all other sites throughout the network. The specific and detailed cost information
for each conceivable rerouting scenario has not, however, been included in the current model
due to time restrictions. However, this first technique offers a fundamental comprehension of
cost dynamics and acts as a starting point for further improvement. The model may be adjusted
to accurately reflect the complicated cost variances that exist across various pathways by
thoroughly gathering and adding precise cost data for each prospective route. Furthermore, a
thorough examination might go beyond particular rerouting pathways to include the whole
spectrum of suitable locations. This entails carefully weighing the costs of delivering items
from one qualifying site to another within the network. The model may obtain a holistic view
of the cost environment by undertaking this in-depth investigation, discovering trends,
anomalies, and significant cost-saving possibilities that might otherwise go unreported. Future
study in this area has the potential to greatly improve the accuracy and robustness of rerouting

44
cost calculations. One area for improvement is determining accurate transportation costs
between specific network points.

Moreover, in the interest of operational safety and risk reduction, when rerouting, a reasonable
methodology might include accounting for the maximum cost associated with each qualifying
site. This preventative strategy ensures that the model is optimised not only for cost efficiency
but also for probable worst-case scenarios. By incorporating this safety margin into the analysis,
the model may provide decision-makers with a more complete picture of anticipated costs and
promote proactive planning to deal with unexpected scenarios.

Assembly Cost

The cost of assembly is a significant factor in establishing a network for a certain order. In
general, one of the locations may be less expensive in terms of transportation and storage, but
the cost savings may be negated by the assembly cost. The availability of workers and
infrastructure is critical in determining assembly costs. Another factor to consider is the
assembly site's learning curve for the specific semiconductor technology. Some assembly sites
might be in the beginning of their learning curve, which results in more rework than locations
near the peak of the learning curve, which results in higher yield. To assess assembly cost, the
yield and other assembly costs must be evaluated simultaneously.

Assembly Site Capacity

The capacities at different downstream partners are assumed to be infinite in this project.
However, in reality, the capacity might be a constraint for certain product lines due to different
processes and requirement needed to manufacture the goods. The assembly sites might also
have their strategic plans that they allocate more resources to certain products or to prioritize
specific businesses and can only process partial of the order. If it happens, the upstream
company faces a higher probability of rerouting goods, and planners need to allocate the order
quantity to the second or even third sources.

Die Bank Capacity

Similar to the assembly site capacity mentioned above, the capacities at all die banks are
assumed to be infinite in this project. Meaning that the warehouses always have sufficient space
and management resources for storing wafers. It might not hold true to all die banks in the

45
supply web, and the capacity can be tight when there are seasonal causes. The planners might
need to consider the logistic plan according to this factor along with other cost related factors.

Although some of the constraints and factors fall into the operations management category
instead of pure supply chain management, those still influence the final decisions and the
fulfilment plans of orders, and further have impacts on how the company design its supply
network ahead, as well as provide the overview of daily order management.

46
References
Abdelshafie, A., Salah, M., Kramberger, T., & Dragan, D. (2022). Repositioning and Optimal Re-
Allocation of Empty Containers: A Review of Methods, Models, and Applications.
Sustainability, 6655.
Analog Devices. (2023). Analog Devices - Analog.com - Analog Devices Solutions. Retrieved 7 8,
2022, from https://www.analog.com
ATN, M. (n.d.). Assembly Process. Retrieved from ALTER: https://wpo-
altertechnology.com/assembly-processes-semiconductor-optoelectronic-microelectronics/
Borgonovo, E. (2017). Sensitivity Analysis: An Introduction for the Management Scientist. Milan:
Springer International Publishing.
Can Sun, T. R. (2015). Supply Chain Complexity in the Semiconductor Industry: Assessment from
System View and the Impact of Changes. IFAC-PapersOnLine, 1210-1215.
Chien, C.-F., Kuo, H.-A., & Lin, Y.-S. (2022). Smart semiconductor manufacturing for pricing,
demand planning, capacity portfolio and cost for sustainable supply chain management.
International Journal of Logistics Research and Applications, 8-10.
Chopra , S., & Sodhi, M. (2004). Managing Risk to Avoid Supply-Chain Breakdown. MIT Sloan
Management Review, 46(1), 53.
Denton, B., Forrest, J., & Milne, R. (2006). IBM Solves a Mixed-Integer Program to Optimize Its
Semiconductor Supply Chain. Interfaces, 386–399.
DiPietro, B. (2015, 10 21). When Manufacturing Means Building Supply-Chain Resilience; For ON
Semiconductor, a risk-management plan was put into motion even as a series of natural
disasters disrupted global supply chains. Retrieved 07 03, 2023, from
https://ucd.idm.oclc.org/login?url=https://www.proquest.com/newspapers/when-
manufacturing-means-building-supply-chain/docview/1724573405/se-2
Entegris. (n.d.). SB300 Full Pitch Front Opening Shipping Boxes (FOSBs). Retrieved 08 02, 2023,
from Entegris: https://www.entegris.com/shop/en/USD/products/wafer-handling/wafer-
shipping/300-mm-wafer-shippers/SB300-Full-Pitch-Front-Opening-Shipping-Boxes-
%28FOSBs%29/p/SB300FullPitchFrontOpeningShippingBoxesFOSBs
EPS. (2011, March 31). Wafer Storage & Management for Extending Product Lifecycle. EPS News,
p. 1.
Ewen, H., Mönch, L., Ehm, H., Ponsignon, T., Fowler, J., & Forstner, L. (2017). A Testbed for
Simulating Semiconductor Supply Chains. IEEE Transactions on Semiconductor
Manufacturing, 293-305.
FedEx. (2023). Calculate FedEx shipping rates. Retrieved 07 27, 2023, from
https://www.fedex.com/en-ie/online/rating.html#
Foreman, J., Bhatnagar, R., Gallien, J., Lopez, F., Alspaugh, J., Teo, C., & Dubois, C. (2010).
Implementing Supply-Routing Optimization in a Make-to-Order Manufacturing Network.
Manufacturing & Service Operations Management, 12(4), 547-568.
Franková, P., Drahošová, M., & Balco, P. (2016). Agile project management approach and its use in
big data management. Procedia Computer Science, 83, 576-583.
Fu, W., Jing, S., Liu, Q., & Zhang, H. (2023). Resilient supply chain framework for semiconductor
distribution and an empirical study of demand risk inference. Sustainability, 15(9), 7382.
Garland, M. (2022). Why parcel shipping will be even more expensive in 2022. Retrieved 08 04, 2023,
from https://www.supplychaindive.com/news/elevated-parcel-shipping-costs-climb-2022-
fedex-ups/617915/
Goodman, J. (1991). Proper Care and Use of Wafer Handling Products. Solid State Technology, 33.

47
Gustavsson, T. (2016). Benefits of Agile Project Management in a Non-Software Development
Context : A Literature Review. Riga: Latvijas Universitate.
Heldman, K., Mangano, V., & Feddersen, B. (2018). Project Management Professional Exam Review
Guide (4th ed.). Indianapolis, Indiana: John Wiley & Sons, Inc.
Highways Today. (2022). The 6 factors affecting Freight Shipping Rates in 2022. Retrieved 08 02,
2023, from https://highways.today/2022/05/27/freight-shipping-rates-2022/
Inanc, O., & Zachariadis, M. (2012). The importance of trade costs in deviations from the law-of-one-
price: estimates based on the direction of trade. Economic Inquiry, 50(3), 667.
Intel. (2023). Intel Plans Assembly and Test Facility in Poland. Retrieved 07 06, 2023, from
https://www.intc.com/news-events/press-releases/detail/1627/intel-plans-assembly-and-test-
facility-in-poland
Ivanov, D., & Dolgui, A. (2019). New disruption risk management perspectives in supply chains:
digital twins, the ripple effect, and resileanness. ScienceDirect, 337-342.
Kan Ji, L. N. (2023, 06 14). Understanding the Semiconductor Supply Chain. Retrieved from
rabobank.com: https://www.rabobank.com/knowledge/s011371708-understanding-the-
semiconductor-supply-chain
Kempf, K. (2004). Control-oriented Approaches to Supply Chain Management in Semiconductor
Manufacturing. American Control Conference (pp. 4563-4576). Boston, MA, USA: IEEE.
Khan, S., Mann, A., & Peterson, D. (2021). The Semiconductor Supply Chain Issue Brief - CSET.
Chicago: Center for Security and Emerging Technology.
Kleindorfer, P., & Saad, G. (2005). Managing Disruption Risks in Supply Chains. Production and
Operations Management, 14(1), 53-68.
Lin, T., Zargar, O., Hu, M., Lin, C., & Leggett, G. (2022). The Effect of Purge Flow Rate and Wafer
Arrangement on Humidity Invasion Into a Loaded Front Opening Unified Pod (FOUP). IEEE
Transactions on Semiconductor Manufacturing, 353-362.
Niemi, T., Hameri, A., Kolesnyk, P., & Appelqvist, P. (2020). What is the Value of Delivering on
Time? Journal of Advances in Management Research, 473-503.
Oliveira, D., & Figueiredo, G. (2014). Determinants of European Freight Rates: The Role of Market
Power and Trade Imbalance. Transportation research. Part E, Logistics and transportation
review, 62, 23-33.
PartnerShip. (2022). Your Essential Guide to the 2023 FedEx and UPS Rate Increases. Retrieved 07
28, 2023, from https://www.partnership.com/blog/post/fedex-ups-general-rate-
increase#:~:text=The%20new%20FedEx%20rates%20take,depend%20on%20several%20diff
erent%20factors.
Place, D. (2018, April 03). SEMIPROBE. Retrieved from Wafer Sort:
https://blog.semiprobe.com/wafer-sort
Politi , M., & Street Jr, R. (2011). The importance of communication in collaborative decisionmaking:
facilitating shared mind and the managementof uncertainty. Journal of Evaluation in Clinical
Practice, 531-838.
Powell, J., Nize, L., & Ji, K. (2023, 06 14). Mapping Global Supply Chains – The Case of
Semiconductors. Retrieved from Rabobank:
https://www.rabobank.com/knowledge/d011371771-mapping-global-supply-chains-the-case-
of-semiconductors
Röhlig Logistics GmbH & Co. KG. (2022). Impact of Russia-Ukraine Conflict on Air Freight.
Retrieved 08 04, 2023, from https://www.rohlig.com/company/blog/impact-of-russia-ukraine-
conflict-on-air-cargo
Rupp, T. (2000). Fine planning for supply chains in semiconductor manufacture. Journal of Materials
Processing Technology, 390-397.

48
Rupp, T., & Ristic, M. (2000). Fine planning for supply chains in semiconductor manufacture.
Journal of Materials Processing Technology, 107(1-3), 390-397.
Saltelli, A., Ratto, M., Andres, T., Campolongo, F., Cariboni, J., Gatelli, D., . . . Tarantola, S. (2007).
Global Sensitivity Analysis. The Primer. London: John Wiley & Sons, Ltd.
Samsung. (2018, May 9). Fabrication Process. Retrieved from Samsung:
https://semiconductor.samsung.com/support/tools-resources/fabrication-process/eight-
essential-semiconductor-fabrication-processes-part-8-eds-electrical-die-sorting-for-the-
perfect-chips/
Scheinberg, N. (2022). FedEx to hike shipping rates by largest increase in at least a decade.
Retrieved 08 04, 2023, from
https://eu.commercialappeal.com/story/money/industries/logistics/2022/09/25/fedex-ground-
express-freight-shipping-rates-increase-costs-2023/69513919007/
Sheffi, Y., & Rice Jr., J. (2005). A Supply Chain View of the Resilient Enterprise. MIT Sloan
Management Review, 47(1), 41.
Silva, M., Fritz, M., & El-Garaihy , W. (2022). Practice theories and supply chain sustainability: a
systematic literature review and a research agenda. emarald Insight, 2631-3871.
Sun, C., & Rose, T. (2015). Supply Chain Complexity in the Semiconductor Industry: Assessment
from System View and the Impact of Changes. IFAC-PapersOnLine, 48(3), 1210-1215.
Sun, Y., Fowler, J., & Shunk, D. (2011). Policies for Allocating Product Lots to Customer Orders in
Semiconductor Manufacturing Supply Chains. Production Planning & Control, 69-80.
Sun, Y., Shunk, D., Fowler, J., & Gel, E. (2010). Strategic Factor-Driven Supply Chain Design for
Semiconductors. California Journal of Operations Management, 8.
Tomlin, B. (2006). On the Value of Mitigation and Contingency Strategies for Managing Supply
Chain Disruption Risks. Management Science, 52(5), 639-657.
United Nations UNCTAD. (2022). THE IMPACT ON TRADE AND DEVELOPMENT OF THE WAR
IN UKRAINE. UNITED NATIONS CONFERENCE ON TRADE AND DEVELOPMENT.
Retrieved from https://unctad.org/system/files/official-document/osginf2022d1_en.pdf
Uzsoy, R., Mönch, L., & Fowler, J. (2018). A survey of semiconductor supply chain models part I:
Semiconductor supply chains, strategic network design, and supply chain simulation.
International Journal of Production Research, 56(13), 4524-4545.
Vietnam Cleanroom. (2022, July 12). The importance of cleanroom in the semiconductor industry.
Vietnam Cleanroom, p. 1.
Wang, W., Rivera, D., & Kempf, K. (2016). Centralized model predictive control strategies for
inventory management in semiconductor manufacturing supply chains. Proceedings of the
2003 American Control Conference (pp. 585-590). Denver: IEEE.
Wanganoo, L., & Shukla, V. (2020). Real-Time Data Monitoring in Cold Supply Chain Through NB-
IoT. International Conference on Computing, Communication and Networking Technologies
(ICCCNT) (p. 6). India: IEEE.
Weiland, T. (n.d.). How Have Fluctuations in Shipping Rates Changed Over Time? Retrieved 08 02,
2023, from https://www.atsinc.com/blog/freight-shipping-costs-over-time
Weizhong, S. (2022). China Patent No. CN202220648642U 20220323 .
Zhou, W., & Lee, C. (2009). Pricing and competition in a transportation market with empty
equipment repositioning. Transportation research. Part B: methodological, 43(6), 677-691.

49