Deriving a global production network type in times of uncertainty – a simulation based approach

Abstract

Global production networks are highly complex to manage and constantly to optimize. Recent developments such as political power changes, pandemic crises or increasing trade hurdles have significantly altered the risk exposure of global production set-ups. We use optimization and simulation tools to derive a suitable network type. We develop a global cross-shipping strategy with an integrated approach combining heuristics and simulation. We quantify the impacts of different uncertainties, such as plant closure and high demand variation with simulation, and it to compare to a local-to-local production network. Our approach makes the model easy to implement and close to real-world processes.
This paper provides support for production network decision-making. We present a scientifically sound and practically feasible approach to an important actual business management problem. The developed integrated approach does not require assumptions about the production network structure or policies and is therefore applicable to a wide range of settings. In our case study, we quantify the positive impact of a global cross-shipping production network in comparison to a local-to-local approach. The result of our study helps to adjust the needed strategic and operational measures to manage a global production network.

Full text

Deriving a global production network type in
times of uncertainty – a simulation based
approach
Shuangqing Liao, Adrian Rüegg, Roman Hänggi
Abstract: Global production networks are highly complex to man-
age and constantly to optimize. Recent developments such as po-
litical power changes, pandemic crises or increasing trade hurdles
have significantly altered the risk exposure of global production
set-ups. We use optimization and simulation tools to derive a suit-
able network type. We develop a global cross-shipping strategy with
an integrated approach combining heuristics and simulation. We
quantify the impacts of different uncertainties, such as plant closure
and high demand variation with simulation, and it to compare to a
local-to-local production network. Our approach makes the model
easy to implement and close to real-world processes.
This paper provides support for production network decision-mak-
ing. We present a scientifically sound and practically feasible ap-
proach to an important actual business management problem. The
developed integrated approach does not require assumptions about
the production network structure or policies and is therefore appli-
cable to a wide range of settings. In our case study, we quantify
the positive impact of a global cross-shipping production network
in comparison to a local-to-local approach. The result of our study
helps to adjust the needed strategic and operational measures to
manage a global production network.
Keywords: Production, production network, strategy, operations,
simulation, decision-making, industrie 4.0, smart factory
Ableitung eines globalen Produktionsnetzwerk-Typs in Zeiten von
Unsicherheit – ein simulationsbasierter Ansatz
Zusammenfassung: Globale Produktionsnetzwerke sind hochkom-
plex zu führen und ständig zu optimieren. Jüngste Entwicklungen
wie politische Machtwechsel, Pandemiekrisen oder zunehmende
Handelshürden haben die Risikoexposition von Produktionsverbünden signifikant ver-
ändert. Wir verwenden Optimierungs- und Simulationswerkzeuge, um einen geeigneten
Netzwerktyp abzuleiten. Wir entwickeln eine globale Cross-Shipping-Strategie mit einem
integrierten Ansatz, der Heuristik und Simulation kombiniert. Zusätzlich quantifizieren
wir die Auswirkungen verschiedener Unsicherheiten, wie Werksschließungen und hohe
Nachfrageschwankungen, mit Hilfe von Simulationen, um die Cross-Shipping-Strategie
552 Die Unternehmung, 75. Jg., 4/2021, DOI: 10.5771/0042-059X-2021-4-552
https://doi.org/10.5771/0042-059X-2021-4-552
Generiert durch Fachhochschule Ostschweiz, am 04.07.2022, 16:06:09.
Das Erstellen und Weitergeben von Kopien dieses PDFs ist nicht zulässig.

mit einem Local-to-Local-Produktionsnetzwerk zu vergleichen. Unser Ansatz macht das
Modell einfach zu implementieren und ist nahe an realen Prozessen.
Diese wissenschaftliche Arbeit bietet Unterstützung für die Entscheidungsfindung in Pro-
duktionsnetzwerken. Wir präsentieren einen wissenschaftlich fundierten und praktisch
umsetzbaren Ansatz für ein wichtiges und aktuelles Problem in der Unternehmensführung.
Der entwickelte integrierte Ansatz erfordert keine Annahmen über die Struktur oder die
Richtlinien des Produktionsnetzwerks und ist daher auf eine breite Palette von Einstel-
lungen anwendbar. In unserer Fallstudie quantifizieren wir die positiven Auswirkungen
eines globalen Cross-Shipping-Produktionsnetzwerks im Vergleich zu einem Local-to-Lo-
cal-Ansatz. Das Ergebnis unserer Studie hilft dabei, die notwendigen strategischen und
operativen Maßnahmen zur Steuerung eines globalen Produktionsnetzwerks anzupassen.
Stichwörter: Produktion, Produktionsnetzwerk, Strategie, Betrieb, Simulation, Entschei-
dungsfindung, Industrie 4.0, Smarte Fabrik
Introduction
Motivation
Most producers today have established several plants abroad and operate in global supply
chains to better address different market needs and optimize cost. Due to this increasing
willingness to outsource steps of the production network on a global scale, the develop-
ment of transport sectors and the setup of bilateral and international trading agreements
have followed suit. In addition, by operating in global production networks, companies
can bypass trade barriers such as customs duties and optimize transport and labor costs
(Friedli et al. 2014). However, the complexity of operating the resulting production net-
works is raised due to the many diverse components and an increasing need of coordina-
tion to deal with costs factors, logistics lead-times, risks and variability.
Given the complex nature of a global production network, it is difficult to manage risks
such as natural disasters, pandemics, political power changes and terrorism. Moreover,
production strategies like lean inventories, just-in-time delivery schedules, centralized dis-
tribution, sourcing from developing countries and global production strategies have been
widely applied. These strategies are most often neither openly debated nor transparent
until an event leads to a supply chain failure. Quantifying the costs and assessing the risks
of a production strategy with a global production network is very difficult due to the
complex interconnection of all stakeholders (Manners-Bell 2017).
Faced to the two main challenges of production network design and operations, com-
plexity and uncertainty, previous research papers have developed sophisticated approaches
such as search algorithms (SA), simulation with design of experiments (DoE), combined
SA and simulation, or closed form solutions for simplified formulations which are far
from real-world processes. However, in many cases, small and medium enterprises (SMEs)
who don't necessarily have capacity in statistics or search algorithms, need fast and
easy-to-implement approaches which achieve sound performance for real-world problems.
This paper responds to this need in decision-making support, by combining heuristics and
simulation. To the best of our knowledge, very little has been done in this area (Paul et al.
2016).
1
1.1
Liao/Rüegg/Hänggi | Deriving a global production network type in times of uncertainty
Die Unternehmung, 75. Jg., 4/2021 553
https://doi.org/10.5771/0042-059X-2021-4-552
Generiert durch Fachhochschule Ostschweiz, am 04.07.2022, 16:06:09.
Das Erstellen und Weitergeben von Kopien dieses PDFs ist nicht zulässig.

In this paper, we use optimization and simulation tools to derive a suitable production
network type. The management decisions are at strategic (e.g., opening new plants) and
operative (e.g., production, procurement and delivery plans) levels. A deterministic mathe-
matical model and a discrete-event simulation (DES) model considering uncertainties have
been developed. First, a heuristic approach that ignores certain constraints is applied to
the mathematical model to obtain an initial strategy and an upper bound on the total
profit. The DES model then improves the initial strategy by considering all constraints and
uncertainties and evaluates the performance of the final strategy with respect to the upper
bound. The simulation model also evaluates the impacts of different uncertainties and the
performances of different production network settings under various conditions.
Our integrated approach improves operational efficiency and provides reasonable mod-
el performance. In our case study, we see that our solution achieves 90 % total profit
comparing to the upper bound obtained. This approach has a high practical contribution
for industrial decision makers. It provides insight into current business management prob-
lems from practice.
Structure of the paper
The remainder of this paper is organized as follows. Section 2 summarizes the state-of-art.
Section 3 introduces the research question, formulates the problem, explains the problem
complexity and presents our integrated approach. Section 4 presents a real-world use case.
Section 5 presents conclusions and findings.
State of the Art
1.2
2
Figure 1: Multi-Supplier Network
Themenbeiträge
554 Die Unternehmung, 75. Jg., 4/2021
https://doi.org/10.5771/0042-059X-2021-4-552
Generiert durch Fachhochschule Ostschweiz, am 04.07.2022, 16:06:09.
Das Erstellen und Weitergeben von Kopien dieses PDFs ist nicht zulässig.

Risk and disruption management have recently become important topics for produc-
tion networks and supply chains. There are many ways to categorize risk types. One
relevant way (Manners-Bell 2017) is to classify risks into internal risks (such as process
and control) and external risks (demand or supply uncertainty, environment disasters,
weather, etc.) Globalization has led to increased risks due to longer lead times, less agile
responses to market conditions, currency fluctuations, labor disputes and shipping costs.
A production network with multiple suppliers (Figure 1) is more complex to manage but
has lower probability of a total disruption compared to a single supplier network (Figure
2). However, as Manners-Bell (2017) stated, “it is very difficult to measure the impact
of an event on a supply chain and even harder to attempt to forecast the impact of a
potential event.”
Solution approaches and research gap
Quantitative approaches solving supply chain network (SCN) problems are mainly divided
into two groups: optimization and simulation.
Optimization methods can be further classified as (i) traditional optimization approach-
es, (ii) heuristic approaches and (iii) search algorithms (metaheuristics). Traditional opti-
mization approaches can solve simple supply chain management problems using methods
such as branch and bound (Timpe/Kallrath 2014), linear programming (Kabak/Ülengin
2011) or quadratic programming (Xia et al. 2004). However, the supply chain network
risk and disruption management problem are usually large-scale, multi-stage and non-lin-
ear. Traditional optimization approaches thus have limits of applicability. Heuristics are
approximate strategies for decision-making and problem-solving that do not guarantee an
optimal solution but that typically yield a reasonable solution (Todd 2001). Heuristics are
simple to understand, easy to apply, and very inexpensive in terms of computing effort,
thus speeding up the process of finding a satisfactory solution (Talbot/Patterson 1979).
Search algorithms use different random-search and parallelization paradigms to obtain
solutions. Examples include genetic algorithms (e.g. Nezamoddini et al. 2020), simulated
annealing (e.g. Diabat 2014), ant colony algorithm (e.g. Bottani et al. 2019), particle
swarm optimization (e.g. Goodarzian et al. 2020), etc.
While optimization methods have been mostly applied to simplified scenarios of real-life
processes, simulation deals with complex processes without mathematical sophistication
but with details and accuracy (Figueira/Almada-Lobo 2014). Simulation is especially suit-
able for complex production networks (Lanza et al. 2019). DES can be defined as an
interacting set of entities that evolve through different states as internal or external events
happen (Robinson 2004). DES is able to produce valid representations of a real system
incorporating the system's uncertainty and dynamics. Semini et al. (2006) reviewed 52
applications of DES to support manufacturing logistic decision-making.
2.1
Figure 2: Single-Supplier Network
Liao/Rüegg/Hänggi | Deriving a global production network type in times of uncertainty
Die Unternehmung, 75. Jg., 4/2021 555
https://doi.org/10.5771/0042-059X-2021-4-552
Generiert durch Fachhochschule Ostschweiz, am 04.07.2022, 16:06:09.
Das Erstellen und Weitergeben von Kopien dieses PDFs ist nicht zulässig.

Simulation is not an optimization tool, however. It needs help from other methods
such as DOE or SA to identify promising solution areas. Hybrid sim-opt methods refer
to the interaction between optimization and simulation “to find near-optimal solutions
to complex or stochastic optimization problems”. A few examples can be found in the
production network management literature. Ding et al. (2009) combined a multi-objective
genetic algorithm (MOGA) and simulation in order to support decisions in the production
distribution network structure and its operation strategies and related control parameters.
DES is used to estimate the operational performance of solutions suggested by the opti-
mizer MOGA. He et al. (2015) dealed with the modeling and optimization problem of
a multi-echelon container supply chain network, where a genetic algorithm (GA) and a
particle swarm optimization (PSO) algorithm are integrated for searching near-optimal so-
lutions, and simulation is used for evaluating solutions and repairing unfeasible solutions.
Aqlan/Lam (2016) proposed an approach combining goal programming and simulation
for supply chain optimization under risk and uncertainty. Chiadamrong/Piyathanavong
(2017) supported decisions for supply chain network design using a combination of ana-
lytical model and DES models, where decisions are split to be separately and iteratively
determined by the two models. Keizer et al. (2015) identifed a cost-optimal network de-
sign under product quality requirements by combining MILP (for strategic optimization)
and simulation (for production quality evalution). Tordecilla et al. (2021) reviewed the
existing literature on the use of simulation-optimization methods in the design of resilient
supply chain networks (SCNs). They stated that applications of hybrid sim-opt methods
are still scarce in the topic of supply chain network design.
Hybrid sim-opt methods usually run the optimization and simulation models iteratively,
solving the analytical part by search or exact algorithms, with parameters or evaluation
results obtained from a simulation model (Juan et al. 2015). In this study, we take a
different approach. We combine a heuristic with simulation, where the heuristic is used
to deterimine an initial strategy and a profit upper bound and the simulation is used to
improve this strategy by taking into account uncertainties and complex constraints. The
upper bound obtained from the heuristic approach serves as a reference for the strategy’s
performance.
Paul et al. (2016) reviewed the mathematical models and the solution approaches used
to solve models for managing risk and disruption in production-inventory systems and
supply chains. They stated that, "It is observed that, most studies focused on using search
algorithm to solve the models. A good number of works also have been found which
developed heuristic and simulation approach to solve the complex models. In case of
dynamic and complex problem, it is worth to develop a combined heuristic and simulation
approach to make the model easy to implement and closer to a real-world process." They
further pointed out that, "Some papers developed a heuristic to solve their models, but
very little has been done to develop a combined heuristic and simulation approach to
operate a model as a real-world process".
Conceptual model
In this paper we analyze a production network design and operation problem. The objec-
tive is to maximize total profit while maintaining a specified on-time-delivery rate (OTD)
and fill rate under certain constraints. Questions to be answered include:
2.2
Themenbeiträge
556 Die Unternehmung, 75. Jg., 4/2021
https://doi.org/10.5771/0042-059X-2021-4-552
Generiert durch Fachhochschule Ostschweiz, am 04.07.2022, 16:06:09.
Das Erstellen und Weitergeben von Kopien dieses PDFs ist nicht zulässig.

§What production network type is proposed? What are the procurement plan, produc-
tion plan and delivery plan over the considered time horizon for this production net-
work?
§Under deterministic /uncertain conditions, how does the performance of the proposed
global production network (where materials and products are distributed across conti-
nents) comparing to the existing local production network?
§What would be the financial impact of an event (uncertainty) on the (proposed) global
vs. (existing) local production networks?
We propose a new solution approach combining a heuristic and simulation. This approach
describes a new way that can be widely used in production network decision-making for
SMEs. It brings a practical contribution to companies facing supply chain disruption risks
and an academic contribution in filling the research gap mentioned above.
This approach is based on in-depth data analysis and process understanding. Prior
to the development of the optimization and simulation model, extensive data collection
took place and valuable insights into the case company were gained. Our work could
benefit from direct access to all relevant data and people in the company. The model was
constantly tested and challenged. This research methodology results in a high degree of
congruence between the simulation results and reality.
Research Design
In this paper we consider a single product production network with multiple globally lo-
cated plants, suppliers and customers. To receive raw materials from suppliers, the plants
pay raw material, transportation and import costs. Each supplier has its own capacity
and commercial limits. Similarly, by delivering product to customers, the plants receive
revenue per batch and pay transportation costs (per shipment) and export taxes. Each
delivery path (supplier-plant or plant-customer) has a maximum volume and a transporta-
tion time. The OTD rate for each customer is required to achieve a target level.
The product is produced and delivered in full batches. Different machine versions are
available which possess different production times (and production costs) per batch,
where the more advanced machine version has higher production priority. Each plant
has its own planned yearly production time, numbers of different machine versions and
overall equipment efficiency (OEE). The capacity utilization of each plant is bounded
above. Production and overhead costs occur in the plants. Inventory costs are included in
the logistic overhead. In each plant, inventory levels of both raw material and product are
required to stay between the available corresponding inventory space and safety stock (SS)
levels.
We now present the full mixed integer program (MIP) formulation of the multi-period
production network. See Appendix I for the model notation.
3
Liao/Rüegg/Hänggi | Deriving a global production network type in times of uncertainty
Die Unternehmung, 75. Jg., 4/2021 557
https://doi.org/10.5771/0042-059X-2021-4-552
Generiert durch Fachhochschule Ostschweiz, am 04.07.2022, 16:06:09.
Das Erstellen und Weitergeben von Kopien dieses PDFs ist nicht zulässig.

Objective function:

,

,


= 1

= 1

= 1
×

−= 1
,, (1)
,, = = 1

= 1
×1 + 

+

+

+= 1

= 1


×× + ×

,
∀
(2)
Suppliers’ capacity and commercial limits:
= 1

= 1
 ≤

,

∀ (3)
= 1

= 1
 ≤

,

∀ (4)


≤= 1

= 1

= 1
,

∀ (5)
Plants’ raw material inventory level constraints:

≤ ≤

,

∀,

 (6)

=

  − 1 − ×

+= 1
 − ,

∀,

 (7)
Plants’ finished product inventory level constraints:

=

  − 1 +  − −= 1
,

∀,

 (8)

=






,

∀ (9)

≤ ≤

,

∀,

 (10)
Plant’s capacity utilization constraints:
∑= 1

∑= 1
 ××

≤

,

∀ (11)
Production and overhead costs:
=1× = 1
,

1××
1
+= 2
 ×

 = 1
 −= 1
− 1  ××

, 0 ,

 ××

× 1 + ,

∀
(12)
Themenbeiträge
558 Die Unternehmung, 75. Jg., 4/2021
https://doi.org/10.5771/0042-059X-2021-4-552
Generiert durch Fachhochschule Ostschweiz, am 04.07.2022, 16:06:09.
Das Erstellen und Weitergeben von Kopien dieses PDFs ist nicht zulässig.

OTD constraints:

1= 1,= 1
 1−

,

∀ (13)

 = , = 1

= 1
 − −= 1
− 1

 , 0

,

∀,

= 2, …,

 (14)
∑= 1



∑= 1

≥,

∀ (15)
Variable restrictions:
,

 ∈0
+,

 ≥ 0,

∀,

,

,

∈ 1,  (16)
The objective is to maximize the total profit over the considered periods, which is the
difference between the total revenues and total costs of all plants. Equation (2) indicates
the three parts of total cost: purchasing, production and delivery costs. Constraints (3)-(5)
enforce that each supplier cannot deliver more than its capacity and commercial limits.
Constraints (6)-(10) require inventory levels to stay between the corresponding inventory
space and SS level. Constraint (11) bounds the capacity utilization of each plant to be
the actual output divided by the potential output of all machine versions. Equation (12)
defines the production and overhead costs in a plant. The machine versions with small
value (v) have higher priority. Equations (13) and (14) define the OTD amount for each
customer at each period: when new batches arrive to a customer in period t, they first fill
the pending demand from previous periods (if it exists) and then fill the demand of the
actual period. Constraint (15) bounds the target OTD level of each customer. Constraint
(16) is the variable constraint. We assume that the initial amounts purchased, produced
and delivered are zero ( = 0,  = 0 , xi = 0, <0 in the periods that are outside
our considered scope.
Note that in order to keep the model a reasonable size, we simplified the product
inventory levels and production costs. In functions (8) and (9), the biggest processing time
among all machines PLT is used to approximate the production time for one batch in
plant j. Similarly, in function (12), the yearly production cost is estimated by assuming
that old version machines work only when new version machines are all fully busy, while
in reality, older machines may continue to work when new machines have completed their
jobs. Thus, the real production cost might be slightly higher than our definition in (12).
We can see that even with the above-mentioned simplifications, it is already very com-
plex to find an optimal solution for our problem. Firstly, it is a multistage dynamic prob-
lem with J×T+J×K×T integer and I×J×T real variables. The state of future stages
depends on the decision variable values in previous stages. Expensive computer processing
capacity and long solution times are necessary if we wish to solve it using a dynamic
programming approach. Furthermore, randomness such as demand variability needs to be
considered. Its size also proves problematic should we wish to use stochastic programming
to search for an optimal solution under uncertainty. Adjustable robust optimization can be
utilized to solve dynamic (multi-stage) production-inventory problems under uncertainty.
For tractability, however, it requires constraints to be linear or convex in the decision
variables (Yanıkoğlu et al. 2019) and constraint (12) doesn't fulfill this requirement.
Liao/Rüegg/Hänggi | Deriving a global production network type in times of uncertainty
Die Unternehmung, 75. Jg., 4/2021 559
https://doi.org/10.5771/0042-059X-2021-4-552
Generiert durch Fachhochschule Ostschweiz, am 04.07.2022, 16:06:09.
Das Erstellen und Weitergeben von Kopien dieses PDFs ist nicht zulässig.

As discussed in Lanza et al. (2019), optimization relies on simplifying assumptions.
Finding optimal solutions is ambitious for real-life problems due to their complexity.
Simulation is a widely used methodology for decision support in global production net-
works, as it facilitates the analysis of the system's behavior under a variety of operating
conditions. However, assuming that we consider n=10 variations of each parameter for
a production network with three suppliers, three plants and three customers, over 52 peri-
ods (weeks), there are n3×52 + 3×3×52 + 3×3×52= 101092 possible combinations. It is simply
not practical to apply Design of Experiments and use pure simulation to get insights for
our problem due to the vast number of parameter combinations.
We propose a combined heuristic and simulation approach to obtain an approximate solu-
tion to our problem. The heuristic model relaxes the requirements of the original optimiza-
tion model and quickly provides a solution as the starting point for the simulation model,
which keeps all settings close to the original problem and improves the solution with itera-
tions. Our approach greatly reduces the solution time but keeps the model close to reality.
Step 1: Approximate the original mathematic model and obtain an initial strategy.
1) The multi-stage problem is transferred into a single stage problem. We search for year-
ly allocation decisions , ∈Z0
+ , xi≥ 0,  ∀i, ,  instead of weekly allocation
decisions as in constraint (16).
2) In constraint (2), 
Vmax
is approximated by 
Vmax
, and all integer constraints in (16)
are relaxed. We also ignore the inventory and OTD constraints, (6)–(10) and (13)–(15).
3) Using Excel solver, we get an optimal solution for the relaxed problem.
4) We find the nearest integer point to the Excel solve solution and use it as the base of
the yearly decisions.
5) Dividing the yearly decision variables by period and rounding when necessary, we get
the weekly decision variables.
All the above simplifications underestimate the total costs of the production network. The
resulting total profit provides an upper bound on the real optimal total profit.
Step 2: Build a DES model considering inventory, OTD and integer constraints and
improve the initial strategy.
We use the program SIMIO (version 12, Vieira et al. 2020) to build a discrete-event sim-
ulation model (DESM1) for our production network. We use the initial strategy from Step
1 as a starting point and develop a local search technique to improve the strategy. In this
step, all constraints which were ignored in Step 1 apply. We improve the cross-shipping
strategy by iteratively adjusting the delivery frequency and amount until the inventory and
OTD constraints are met. The total profit upper bound from Stage 1 serves as a reference
to judge the performance of the improved cross-shipping strategy.
Step 3: Add uncertainty to our simulation model and add adaptation measures to the
cross-shipping strategy.
To the discrete-event simulation model (DESM2) for our production network, we add
three types of uncertainties: demand variation, supplier failure and plant breakdown. For
each type of uncertainty, we propose adaptation measures to the cross-shipping strategy
from Step 2.
Concerning demand variations, we assume that the weekly demands of all customers vary
with a certain percentage, but the total yearly demands keep constant. The customers
Themenbeiträge
560 Die Unternehmung, 75. Jg., 4/2021
https://doi.org/10.5771/0042-059X-2021-4-552
Generiert durch Fachhochschule Ostschweiz, am 04.07.2022, 16:06:09.
Das Erstellen und Weitergeben von Kopien dieses PDFs ist nicht zulässig.

communicate their variation in advance (frozen window, e.g., 4 weeks) and the plants adjust
their original production plan accordingly. We add the following adaptation measures:
§Delivery plan: At the beginning of each week, the expected OTD rate of each customer
in the next frozen window time is calculated. If it is lower than 100 %, the nearest plant
delivers the missed batches just-in-time using backward scheduling.
§Production plan: The production plan of the plant which has delivered the missed
batches is adjusted accordingly.
§Purchasing plan: No change.
Concerning supplier failure, we assume different failure rate for each supplier. We add the
following adaptation measures:
§Delivery plan: When a small supply failure occurs, the corresponding plants reduce
their delivery amount for their farthest customer and the missed amount for the
customer is delivered by a local plant. If the supplier failure rate is rather high, the
corresponding plants only deliver the portion that they can. The other plants produce
the missed batches and deliver to all customers.
§Production plan: The production plan is adapted to the delivery plan accordingly.
§Purchasing plan: The other suppliers provide the needed material accordingly.
Plant breakdown: Inspired by the actual COVID-19 pandemic, plant breakdown is consid-
ered by decreasing plant availability within a defined time frame. The adaptation measures
use the same rules as with supplier failures.
Using simulation, we then evaluate the performance of the above developed cross-ship-
ping comparing to other strategies (e.g., local-to-local strategy) and gain useful insights.
Mathematic
Model (MM)
Relaxed
Mathematic
Model (RMM)
Discrete
Event
Simulation
Model 1 (DESM1)
Cross-Shipping-
Strategy
Optimal Solution to
the RMM Problem for
the Cross-Shipping-
Strategy (initial Strategy)
Improve strategy
considering constraint s
Discrete
Event
Simulation
Model 2
(DESM2)
Adding Un-
certainties
Adding measure to
operationalize strategy
Local-to-Local-
Strategy
Performance
evaluation of
two different
strategies
Total profit of
the production
network
based on
DESM1
Upper bound
total profit of
the production
network based
on RMM
Evaluation of Cross-Shipping-Strategy
Comparison
Reduction of Model
Complexity by
ignoring constraints
Step 1
Step 2 Step 3
Strategy development
Strategy Comparison
Figure 3: Overview of our approach
Liao/Rüegg/Hänggi | Deriving a global production network type in times of uncertainty
Die Unternehmung, 75. Jg., 4/2021 561
https://doi.org/10.5771/0042-059X-2021-4-552
Generiert durch Fachhochschule Ostschweiz, am 04.07.2022, 16:06:09.
Das Erstellen und Weitergeben von Kopien dieses PDFs ist nicht zulässig.

Figure 3 provides an overview of our approach with two different focus areas. One focus
area shows the development steps of the cross-shipping strategy, as explained above. The
second focus area features a comparison of the strategies' performances. As mentioned
above, the total profit upper bound obtained from the heuristic approach (Step 1) is used
as a reference to evaluate the performance of the improved strategy (Step 2).
Case Study
The developed approach in Section 3 is applied in a SME. A product is produced in sever-
al plants located in the US and Europe following the same production process with three
different versions of machines available. Raw materials are provided by three suppliers,
one in the US and two in Europe. Three customers are demanding products, two in the US
and one in Europe. We refer to "local-to-local" vs. "cross-shipping" production networks
depending on whether materials and products are distributed within or across continents.
We build a simulation model as in Figure 4, using all the parameters and constraints
described in Section 3. KPIs such as revenues, costs, OTD rate, and capacity utilization are
displayed numerically. The inventory levels comparing inventory spaces and SS levels are
displayed graphically (see Appendix II-IV).
Figure 4: Simulation model
We compare the performance of two networks under various conditions (scenarios): the
actual network with two sites and a local-to-local supply strategy (Figure 5) and a global
network with three sites and a global cross-shipping supply strategy (Figure 6). For the
local-to-local strategy, each plant uses the backward scheduling method (just-in-time) to
deliver to its corresponding customers, produce the exact required amount, and purchase
raw material according to the production plan. The cross-shipping strategy is developed as
in Section 3.
4
Themenbeiträge
562 Die Unternehmung, 75. Jg., 4/2021
https://doi.org/10.5771/0042-059X-2021-4-552
Generiert durch Fachhochschule Ostschweiz, am 04.07.2022, 16:06:09.
Das Erstellen und Weitergeben von Kopien dieses PDFs ist nicht zulässig.

Figure 5: Local-to-local network Figure 6: Global cross-shipping network
In the first scenario, all parameters are deterministic. Weekly demand from each customer
is assumed to be stable and known from the beginning of the year. In the next three
scenarios, demand variation, supplier failure and plant breakdown are considered with
the cross-shipping strategy applying the adaptation measures (presented in Section 3)
accordingly. In Scenario 2, the local-to-local network does not apply a supply-chain adap-
tation since the transport time is shorter than the frozen window. In Scenario 3, different
supplier failure rates have been applied for the supplier in the US. Under this condition,
the local-to-local network can't make any adaptation since the US-plant relies on its
only supplier located in the US. In Scenario 4, the US-production plant availability has
been reduced. In this case, the local-to-local network adapts the raw material purchasing
amount according to the new plant availability.
Results
In this section we discuss the results of the four scenarios simulated.
0%
5%
10%
15%
20%
25%
30%
35%
Material cost
Production cost
Transport cost
Overhead cost
Profit
Local network Global network
Cost and Profit [%]
Figure 7: Costs and Profits local vs. global network for Scenario 1
4.1
Liao/Rüegg/Hänggi | Deriving a global production network type in times of uncertainty
Die Unternehmung, 75. Jg., 4/2021 563
https://doi.org/10.5771/0042-059X-2021-4-552
Generiert durch Fachhochschule Ostschweiz, am 04.07.2022, 16:06:09.
Das Erstellen und Weitergeben von Kopien dieses PDFs ist nicht zulässig.

In Scenario 1, both local and global networks achieve 100 % OTD and fill rates. Within
the local set up, both plants’ utilizations are under 75 %. Within the global network, all
three plants are well below 80 % utilization. A financial comparison (Figure 7) shows that
the global network has a 2.2 % higher profit margin than the local setup, driven by lower
production and overhead costs, but higher material costs.
It is worth mentioning that the total profit of our global cross-shipping network based
on the discrete-event simulation model is 90 % of the total profit upper bound we get
from the relaxed mathematical model in Section 3, Step 1. This upper bound provides a
reference for judging how well a strategy behaves financially.
-3%
-2%
-1%
0%
1%
2%
3%
0% 50% 100% 150% 200% 250% 300%
Global
Network
Local Network
Profit Delta [%]
Variation [%]
Figure 8: Profit vs. Demand Variation for Scenario 2
Scenario 2 simulates demand variations from 50 % to 300 %. Within the local network
they are compensated with underutilized, older and more expensive machine versions.
Whereas, in the global network, demand variations are compensated by cross-shipping
within the network. Demand variation has more significant negative impacts on profit for
both local and global networks (Figure 8). Deliveries start to be late within the global
30%
40%
50%
60%
70%
80%
90%
100%
110%
0% 50% 100% 150% 200% 250% 300%
Customer Europe 1 (local)
Customer USA 1 (local)
Customer USA 2 (local)
Customer Europe 1 (global)
Customer USA 1 (global)
Customer USA 2 (global)
Variation [%]
On time delivery [%]
Figure 9: OTD vs. Demand Variation for Scenario 2
Themenbeiträge
564 Die Unternehmung, 75. Jg., 4/2021
https://doi.org/10.5771/0042-059X-2021-4-552
Generiert durch Fachhochschule Ostschweiz, am 04.07.2022, 16:06:09.
Das Erstellen und Weitergeben von Kopien dieses PDFs ist nicht zulässig.

network at a demand variation of 100 % (Figure 9). However, the OTD rate is only
slightly impacted up to a demand variation of 200 % and concerns only one customer site.
On the other hand, the local network's OTD rate falls steeply from variations over 150 %.
We can conclude that, in our case, the global network is more stable in terms of OTD
under demand variation.
Generally, both networks could achieve a 100 % fill rate at demand variation of 300 %.
Only at 300 % demand variation within the global set up does the fill rate fall slightly
below target at 97 %.
Except for a few negligible SS violations, product inventories for the local network are kept
well within the required borders below 100 % demand variation. Within the global network,
product is kept to acceptable levels up to 150 % variation. At higher variations product
stocks fall critically low for both networks. Conversely, with quite small variations, raw
material storage gets very low (global set up) and extremely high (local set up). This could be
improved by adjusting the purchasing plan (which was not a part of this model). An example
of the inventory development can be found in Appendix II.
In Scenario 3, supplier failures and raw material rejections of the US-raw material
supplier were simulated. As rejected raw material is not paid, and raw material stocks are
just consumed, the total profit of the local network increases for failure rates of up to
15 % (Figure 10). At higher failure rates the profit decreases as the fill rates decrease and
fewer products are sold (Figure 12). Consequently, deliveries are late and the OTD rate
drops (Figure 11).
-1%
0%
1%
2%
3%
0% 15% 30% 50%
Global Network
Local Network
Supplier Failure [%]
Profit Delta [%]
Figure 10: Profit vs. Supplier Failure Rate for Scenario 3
The global network is more affected at failure rates below 15 %. Profits decrease (Figure
10), deliveries are late (Figure 11) and demands cannot be fulfilled (Figure 12). Increasing
failure rates from 15 % to 50 %, demonstrates the strategy adaptation’s positive impact.
The plant Europe 1 takes over short-delivered orders from the US-plant. The US-plant
only executes local deliveries. With changed strategies the total profit increases as the
failure rate goes up to 30 % from 15 % (Figure 10). In the worst case, at a 50 %
failure rate, the profit is less impacted than that of the local network. The OTD rates
are stabilizing (Figure 11) except for customer Europe 1. More important is the fill rate
development (Figure 12). At a failure rate of 30 %, two of three customers still receive the
full demand. Furthermore, the overall fill rate at failure rates larger than 30 % is better
than with the local network set up.
Liao/Rüegg/Hänggi | Deriving a global production network type in times of uncertainty
Die Unternehmung, 75. Jg., 4/2021 565
https://doi.org/10.5771/0042-059X-2021-4-552
Generiert durch Fachhochschule Ostschweiz, am 04.07.2022, 16:06:09.
Das Erstellen und Weitergeben von Kopien dieses PDFs ist nicht zulässig.

10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
110%
0% 15% 30% 50%
Customer Europe 1 (local)
Customer USA 1 (local)
Customer USA 2 (local)
Customer Europe 1 (global)
Customer USA 1 (global)
Customer USA 2 (global)
On time Delivery [%]
Supplier Failure [%]
Figure 11: OTD vs. Supplier Failure Rate for Scenario 3
In summary, at low failure rates the local set up is better in terms of profit, OTD and fill
rates. At higher failure rates the global network absorbs some of the financial loss and can
flatten the negative OTD and fill rate trends.
The inventories at a failure rate of 50 % are presented in Appendix III. Within the
local strategy, the raw material and product at the US-plant are just run to zero. There
is no possibility to take counter measures to avoid a supply shortage. Comparatively, in
the global network set up, by adapting delivery, production and procurement strategies,
the product stocks in the US-plant can be kept on an acceptable level. Taking over the
deliveries of the disturbed US-plant causes product stocks at plant Europe 1 to run at a
low but stable level. The temporarily very high raw material stock can improve with a
better procurement strategy. Hence, a smaller order lot size or longer delivery intervals
should be applied.
70%
75%
80%
85%
90%
95%
100%
105%
0% 15% 30% 50%
Customer Europe 1 (local)
Customer USA 1 (local)
Customer USA 2 (local)
Customer Europe 1 (global)
Customer USA 1 (global)
Customer USA 2 (global)
Fill Rate [%]
Supplier Failure [%]
Figure 12: Fill Rate vs. Supplier Failure Rate for Scenario 3
Themenbeiträge
566 Die Unternehmung, 75. Jg., 4/2021
https://doi.org/10.5771/0042-059X-2021-4-552
Generiert durch Fachhochschule Ostschweiz, am 04.07.2022, 16:06:09.
Das Erstellen und Weitergeben von Kopien dieses PDFs ist nicht zulässig.

Scenario 4 simulates plant breakdown with different levels. The adaptation rules for the
cross-shipping strategy in Scenario 4 are very similar to those in Scenario 3 (as described
in Section 3). Scenario 4 thus shows results similar to Scenario 3. For same reasons,
the local network shows an increased profit at lower breakdown rates and then steadily
decreases (Figure 13).
-14%
-12%
-10%
-8%
-6%
-4%
-2%
0%
2%
4%
0% 15% 30% 50% 70% 100%
Global Network
Local Network
Profit Delta [%]
Plant Breakdown [%]
Figure 13: Profit vs. Plant Breakdown for Scenario 4
As we have simulated a wider range in this scenario, the higher robustness of the global
network regarding profit is more obvious. At breakdown rates higher than 30 % the
global network results in less financial loss. Furthermore, the OTD performance of the
global network is better. For example, at a breakdown rate of 70 % the global network
still on average delivers 62 % of the parts on time, whereas only 46 % of parts are on time
within the local set up (Figure 14).
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
110%
0% 15% 30% 50% 70% 100%
Customer Europe 1 (local)
Customer USA 1 (local)
Customer USA 2 (local)
Customer Europe 1 (global)
Customer USA 1 (global)
Customer USA 2 (global)
Plant Breakdown [%]
Delivery on time [%]
Figure 14: OTD vs. Plant Breakdown for Scenario 4
Crucial is the fill rate, which indicates how much demand could be fulfilled at the end
of the simulation period (Figure 15). Between a 30 % to 70 % breakdown rate the
global network significantly outperforms the local network. For example, with a 70 %
breakdown rate, 85 % of the products can still be delivered, whereas the local network
correspondingly delivers only 68 %.
Liao/Rüegg/Hänggi | Deriving a global production network type in times of uncertainty
Die Unternehmung, 75. Jg., 4/2021 567
https://doi.org/10.5771/0042-059X-2021-4-552
Generiert durch Fachhochschule Ostschweiz, am 04.07.2022, 16:06:09.
Das Erstellen und Weitergeben von Kopien dieses PDFs ist nicht zulässig.

10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
110%
0% 15% 30% 50% 70% 100%
Customer Europe 1 (local)
Customer USA 1 (local)
Customer USA 2 (local)
Customer Europe 1 (global)
Customer USA 1 (global)
Customer USA 2 (global)
Fill Rate [%]
Plant Breakdown [%]
Figure 15: Fill Rate vs. Plant Breakdown for Scenario 4
Concerning the inventories, a similar development is seen as in Scenario 3 (Appendix IV).
In both scenarios, the inventories of the disturbed plant (the US-plant) are running low.
They recover at the end of the simulation period in the global network.
Similar statements can be made as in scenario 3. The global network is more sensitive
in terms of profits, OTD and fill rates at breakdown rates up to 15 %. Its performance
increases with higher breakdown rates from 30 % to 70 %. At a 100 % breakdown rate
both networks show poor OTD and fill rate performance but the financial loss within the
global set up is smaller.
In summary, within the settings of our use-case, the local network has advantages when
disruptions are small while the global network performs better under larger disruptions.
With more sophisticated procurement, production and delivery strategies, the performance
of the global production network could be further improved. Improvement potentials
within the local network to counter steer the impacts of uncertainties are limited.
Conclusion
In this work, we consider a complex multi-stage production network problem, originating
in a real case of a Swiss SME. The objective is to maximize the total profit of the
whole production network with respect to various constraints such as OTD rate, fill rate,
inventory levels, and capacities. We have developed a new integrated approach combining
heuristics and simulation to solve the dynamic, complex, large-scale MIP problem under
uncertainty. This approach can be widely and generically used for operational problems in
production networks and supply chain management.
In our case-study, we develop a global cross-shipping strategy using optimization and
simulation tools. With the developed simulation model, under various conditions, we
evaluate the performance of the new cross-shipping strategy as compared to the actual
local-to-local strategy. The results in our case study show that, within our problem set-
ting, a global network is less vulnerable to selected uncertainties. For large variations
of uncertainties, the global network outperforms (in profit, OTD and fill rate) the local
network setup in all scenarios. Further development of optimal strategies and timings for
strategy adaptions could even further improve the results of the global network set up. An
5
Themenbeiträge
568 Die Unternehmung, 75. Jg., 4/2021
https://doi.org/10.5771/0042-059X-2021-4-552
Generiert durch Fachhochschule Ostschweiz, am 04.07.2022, 16:06:09.
Das Erstellen und Weitergeben von Kopien dieses PDFs ist nicht zulässig.

interlinked, cross-shipping global network with multiple suppliers, production plants and
customers offers many more possibilities regarding operational adjustments in uncertain
times.
We emphasize that the above discovery is based on the parameter settings of our specific
case study. It doesn't necessarily generalize. However, the integrated approach that we
developed can be widely used to gain insights for other production networks. We believe
that decision-makers in industry could significantly benefit from this easy-to-implement
approach to making prompt and reasonable decisions for constantly optimizing produc-
tion networks at a strategic and operational level. At the same time, our work fills the
research gap of developing combined heuristic and simulation approaches with models
that are both easy to implement and close to real-world processes (Paul et al. 2016), not
only in production networks but also in supply chain optimization research.
Appendix
Notations
Parameters
Index
= 1, …,  Supplier
= 1, …,  Plant
= 1, …,  Customer
= 1, …,  Time period (week)
= 1, …,  Machine version
Purchasing
 Capacity limit of supplier i
 Import tax (%) when plant j purchases from supplier i
 Transportation time from supplier i to plant j
 Raw material price per ton from supplier i
 Transportation cost (%) from supplier i to plant j
 Maximum volume (tons) per delivery from supplier i to plant j
 Maximum commercial share for supplier i
Production
b Availability factor in plant j
 Amount of machines version v in plant j
BOM Raw material (tons) needed to produce one batch of product
0 Initial raw material inventory level in plant j
0 Initial product inventory level in plant j
 Overhead costs (production, development, administration, logistic over-
head and sales, %) in plant j
 Production cost for one batch of product made by machine version v in
plant j
f Performance factor in plant j
 Product space in plant j
j Product SS level in plant j
 Planned production time over one year in plant j
 Quality factor in plant j
 Raw material space in plant j
6
I
Liao/Rüegg/Hänggi | Deriving a global production network type in times of uncertainty
Die Unternehmung, 75. Jg., 4/2021 569
https://doi.org/10.5771/0042-059X-2021-4-552
Generiert durch Fachhochschule Ostschweiz, am 04.07.2022, 16:06:09.
Das Erstellen und Weitergeben von Kopien dieses PDFs ist nicht zulässig.

Parameters
 Raw material SS level in plant j
 Production time to produce one batch, by machine version v
 Capacity utilization of plant j
Delivery
 Sale price per batch to customer k
 The demand from customer k at period t
 Shipment cost per delivery from plant j to customer k
 Export tax (%) paid when plant j delivers to customer k
 Transportation time from plant j to customer k
 Maximum volume (batches) per delivery from plant j to customer k
 Target OTD rate of customer k
Variables
 Commercial limit of supplier i
 Total cost for plant j
 OTD amount for customer k in period t
 Inventory level of raw material in plant j at the beginning of period t
 Inventory level of product in plant j at the beginning of period t
 Maximum production time per batch in plant j
 Production and overhead costs for plant j
 The raw material amount (tons) delivered from supplier i to each plant j
 The production amount (batches) at plant j
 The product amount (batches) delivered from plant j to customer k
Themenbeiträge
570 Die Unternehmung, 75. Jg., 4/2021
https://doi.org/10.5771/0042-059X-2021-4-552
Generiert durch Fachhochschule Ostschweiz, am 04.07.2022, 16:06:09.
Das Erstellen und Weitergeben von Kopien dieses PDFs ist nicht zulässig.

Inventory Demand variationII
Liao/Rüegg/Hänggi | Deriving a global production network type in times of uncertainty
Die Unternehmung, 75. Jg., 4/2021 571
https://doi.org/10.5771/0042-059X-2021-4-552
Generiert durch Fachhochschule Ostschweiz, am 04.07.2022, 16:06:09.
Das Erstellen und Weitergeben von Kopien dieses PDFs ist nicht zulässig.

Inventory Supplier failureIII
Themenbeiträge
572 Die Unternehmung, 75. Jg., 4/2021
https://doi.org/10.5771/0042-059X-2021-4-552
Generiert durch Fachhochschule Ostschweiz, am 04.07.2022, 16:06:09.
Das Erstellen und Weitergeben von Kopien dieses PDFs ist nicht zulässig.

Inventory Plant failureIV
Liao/Rüegg/Hänggi | Deriving a global production network type in times of uncertainty
Die Unternehmung, 75. Jg., 4/2021 573
https://doi.org/10.5771/0042-059X-2021-4-552
Generiert durch Fachhochschule Ostschweiz, am 04.07.2022, 16:06:09.
Das Erstellen und Weitergeben von Kopien dieses PDFs ist nicht zulässig.

References
Aqlan, F./Lam, S. S. (2016): Supply chain optimization under risk and uncertainty: A case study for
high-end server manufacturing, in: Computers & Industrial Engineering, Vol. 93, pp. 78–87.
Bottani, E., et al. (2019): Resilient food supply chain design: Modelling framework and metaheuris-
tic solution approach, in: Computers & Industrial Engineering, Vol. 135, pp. 177–198.
Chiadamrong, N./Piyathanavong, V. (2017): Optimal design of supply chain network under uncer-
tainty environment using hybrid analytical and simulation modeling approach, in: Journal of
Industrial Engineering International, Vol. 13, No. 4, pp. 465–478.
Diabat, A. (2014): Hybrid algorithm for a vendor managed inventory system in a two-echelon
supply chain, in: European Journal of Operational Research, Vol. 238, No. 1, pp. 114–121.
Ding, H., et al. (2009): Stochastic multi-objective production-distribution network design using
simulation-based optimization, in: International Journal of Production Research, Vol. 47, No. 2,
pp. 479–505.
Figueira, G./Almada-Lobo, B. (2014): Hybrid simulation–optimization methods: A taxonomy and
discussion, in: Simulation modelling practice and theory, Vol. 46, pp. 118–134.
Friedli, T., et al. (2014): Strategic management of global manufacturing networks: Aligning strategy,
configuration, and co-ordination.
Goodarzian, F., et al. (2020): A multi-objective pharmaceutical supply chain network based on a
robust fuzzy model: A comparison of meta-heuristics, in: Applied Soft Computing, Vol. 92,
pp. 106331.
He, J., et al. (2015): Simulation-based heuristic method for container supply chain network opti-
mization, in: Advanced Engineering Informatics, Vol. 29, No. 3, pp. 339–354.
Juan, A. A., et al. (2015): A review of simheuristics: Extending metaheuristics to deal with stochastic
combinatorial optimization problems, in: Operations Research Perspectives, Vol. 2, pp. 62–72.
Kabak, Ö./Ülengin, F. (2011): Possibilistic linear-programming approach for supply chain network-
ing decisions, in: European Journal of Operational Research, Vol. 209, No. 3, pp. 253–264.
Keizer, M. de, et al. (2015): Hybrid optimization and simulation to design a logistics network for
distributing perishable products, in: Computers & Industrial Engineering, Vol. 88, pp. 26–38.
Lanza, G., et al. (2019): Global production networks: Design and operation, in: CIRP Annals, Vol.
68, No. 2, pp. 823–841.
Manners-Bell, J. (2017): Supply chain risk: Understanding emerging threats to global supply chains.
Nezamoddini, N., et al. (2020): A risk-based optimization framework for integrated supply chains
using genetic algorithm and artificial neural networks, in: International Journal of Production
Economics, Vol. 225, p. 107569.
Paul, S. K., et al. (2016): Managing risk and disruption in production-inventory and supply chain
systems: A review, in: Journal of Industrial and Management Optimization, Vol. 12, No. 3,
pp. 1009–1029.
Robinson, S. (2004): Simulation: The Practice of Model Development and Use, Chichester.
Semini, M./Fauske, H./Strandhagen, J. (Hrsg.) (2006): Applications of Discrete-Event Simulation to
Support Manufacturing Logistics Decision-Making: A Survey.
Talbot, F. B./Patterson, J. H. (1979): Optimal methods for scheduling projects under resource con-
straints, in: Project Management Quarterly, Vol. 10, No. 4, pp. 26–33.
Timpe, C. H./Kallrath, J. (2014): Optimal planning in large multi-site production networks, in:
European Journal of Operational Research, Vol. 126, pp. 422–435.
Themenbeiträge
574 Die Unternehmung, 75. Jg., 4/2021
https://doi.org/10.5771/0042-059X-2021-4-552
Generiert durch Fachhochschule Ostschweiz, am 04.07.2022, 16:06:09.
Das Erstellen und Weitergeben von Kopien dieses PDFs ist nicht zulässig.

Todd, P. M. (2001): Heuristics for Decision and Choice, in: International Encyclopedia of the Social
& Behavioral Sciences, pp. 6676–6679.
Tordecilla, R. D., et al. (2021): Simulation-optimization methods for designing and assessing re-
silient supply chain networks under uncertainty scenarios: A review, in: Simulation modelling
practice and theory, Vol. 106, p. 102166.
Vieira, A. A., et al. (2020): Are Simulation Tools Ready For Big Data? Computational Experiments
with Supply Chain Models Developed in Simio, in: Procedia Manufacturing, Vol. 42, pp. 125–
131.
Xia, Y., et al. (2004): Real-time disruption management in a two-stage production and inventory
system, in: IIE Transactions, Vol. 36, No. 2, pp. 111–125.
Yanıkoğlu, İ., et al. (2019): A survey of adjustable robust optimization, in: European Journal of
Operational Research, Vol. 277, No. 3, pp. 799–813.
Shuangqing Liao, M.Sc., Ph.D. is a senior research assistant and a lecturer of production
management at OST – Eastern Switzerland University of Applied Sciences.
Address: OST – Eastern Switzerland University of Applied Sciences, Oberseestrasse 10,
8640 Rapperswil, Switzerland, Phone: + 41 58 257 46 11, E-mail: shuangqing.liao@ost.ch
Adrian Rüegg, is a B.Sc. student in industrial Engineering at OST – Eastern Switzerland
University of Applied Sciences and Global Head of Product and Process Engineering at the
industrial partner company.
Address: OST – Eastern Switzerland University of Applied Sciences, Oberseestrasse 10,
8640 Rapperswil, Switzerland, Phone: + 41 58 257 46 11, E-mail: adrian.rueeg@ost.ch
Roman Hänggi, dipl. Ing. ETH, Dr. Oec. HSG, is a professor in production management
at OST – Eastern Switzerland University of Applied Sciences.
Address: OST – Eastern Switzerland University of Applied Sciences, Oberseestrasse 10,
8640 Rapperswil, Switzerland, Phone: +41 58 257 46 03, E-mail: roman.haenggi@ost.ch
Liao/Rüegg/Hänggi | Deriving a global production network type in times of uncertainty
Die Unternehmung, 75. Jg., 4/2021 575
https://doi.org/10.5771/0042-059X-2021-4-552
Generiert durch Fachhochschule Ostschweiz, am 04.07.2022, 16:06:09.
Das Erstellen und Weitergeben von Kopien dieses PDFs ist nicht zulässig.