Journal of Intelligent Manufacturing (2021) 32:1899–1911

https://doi.org/10.1007/s10845-020-01724-5

A digital twin emulator of a modular production system using

a data‑driven hybrid modeling and simulation approach
Konstantinos Mykoniatis1 · Gregory A. Harris1

Received: 28 May 2020 / Accepted: 3 December 2020 / Published online: 5 January 2021
© The Author(s), under exclusive licence to Springer Science+Business Media, LLC part of Springer Nature 2021

Abstract
Virtual commissioning is a key technology in Industry 4.0 that can address issues faced by engineers during early design
phases. The process of virtual commissioning involves the creation of a Digital Twin—a dynamic, virtual representation
of a corresponding physical system. The digital twin model can be used for testing and verifying the control system in a
simulated virtual environment to achieve rapid set-up and optimization prior to physical commissioning. Additionally, the
modular production control systems, can be integrated and tested during or prior to the construction of the physical system.
This paper describes the implementation of a digital twin emulator of an automated mechatronic modular production system
that is linked with the running programmable logic controllers and allow for exchanging near real-time information with
the physical system. The development and deployment of the digital twin emulator involves a novel hybrid simulation- and
data-driven modeling approach that combines Discrete Event Simulation and Agent Based Modeling paradigms. The Digital
Twin Emulator can support design decisions, test what-if system configurations, verify and validate the actual behavior of
the complete system off-line, test realistic reactions, and provide statistics on the system’s performance.

Keywords Digital twin · Hybrid simulation · Discrete event simulation · Agent based modeling · Emulator · Modular
production · Automation

Introduction operations. Verification and Validation (V&V) of equip-

ment controlling code is business and safety critical and
Manufacturing systems today are complex and composed of must occur in a manner that protects people, products, and
automated workstations that must contend with increasing equipment. It is important to take into consideration poten-
product variety, varying demand, and shorter time to market tial evolutionary developments of the system and ensure cor-
requirements. Industry must have faster, more cost-effective, rectness, consistency, comprehensibility, and maintainability
reliable, and flexible production capabilities to respond to of the code for several years (Beckert et al. 2019).
this increasing system complexity and variation. One The effort of developing, verifying, and validating an
method of managing this increasing variation and demand automated manufacturing system requires an in-depth
is to develop highly efficient automated production systems understanding of software, hardware, mechanical, electrical,
with the ability to detecting system errors and issues early pneumatic, and hydraulic systems. Additionally, knowledge
in the design and development phases. and insight into how operators engage with these systems is
Currently, such automated production systems are con- critical. Therefore, developers from various disciplines must
trolled by Programmable Logic Controllers (PLCs) that run collaborate to accomplish the desired system behavior in a
advanced logic controller code. PLC code becomes chal- timely manner.
lenging to verify and validate especially as the length of the One method to test PLC code is to debug it manually, or
program grows and the PLCs are stretched to control more by trial and error on the real equipment, which can dam-
age the equipment or harm personnel (Makris et al. 2012).
* Konstantinos Mykoniatis Moreover, the mechanical design and testing, as well as the
[email protected] PLC code debugging and station commissioning, are tra-
ditionally performed sequentially and online, which leads
1
Department of Industrial and Systems Engineering, Auburn
University, 3312 Shelby Center, Auburn, AL 36849, USA

13
Vol.:(0123456789)
1900 Journal of Intelligent Manufacturing (2021) 32:1899–1911

to long set-up times and increased cost (Bathelt and Meile the complete system off-line, tests realistic reactions, and
2007; Pellicciari et al. 2009). provide statistics of the system performance; (3), the devel-
A more efficient way to overcome these challenges is opment of a PLC Tag Synchronizer program for coordinat-
through simulation-based emulation, which is commonly ing high speed Ethernet/IP communication and real time
referred to as Virtual Commissioning (VC). VC provides an data collection and utilization between the DTE and the
avenue for V&V during the development of an automated actual MPS asset; and (4) compatibility features that allow
production system in a lower risk environment (Lee and the DTE to be used for both, Hardware-In-the-Loop (HIL)
Park 2014). Engineers and programmers can virtually test commissioning, when it is connected to the actual PLCs of
PLC code, alternative production designs or concepts, and the Modular Production System (MPS), as well as for Soft-
rework issues prior to the installation of physical equipment. ware-In-the-Loop (SIL) commissioning when it is linked
In addition, VC allows engineers and programmers to work to a virtual control system and virtual fieldbus emulators.
concurrently and in parallel (Carlsson et al. 2012). This way, The remainder of the paper is organized as follows:
they can virtually verify and test their programs and rapidly Section “Background” provides an overview of existing
set-up and optimize a production system early in the design commissioning approaches and Digital Twins (DTs). The
phase thus reducing development time, costs and risks, and Section “Methodology of digital twin emulator for virtual
improving competitiveness (Thramboulidis 2005; Hoffmann commissioning of the festo modular production system”
et al. 2010; Andersen et al. 2016). describes the methodology that we followed for the imple-
In this paper, we describe a novel hybrid simulation- and mentation of the hybrid DES-ABM architecture of the DTE.
data-driven Digital Twin Emulator (DTE) for commission- Section “Communication and time synchronization between
ing an automated mechatronic Modular Production System the PLCs and the DTE” describes the communication and
(MPS). Although the idea of the Digital Twin (DT) has not time synchronization between the PLCs and the DTE, Sec-
changed since the basic concept was conceived around 2002 tion “Experimentation and results” presents the results
(Shao et al. 2019), it is generally considered to be a digital and evaluation of the developed DTE. Finally, in Section
informational construct that mirrors the behavior of a physi- “Conclusions and future work” we discuss future work and
cal system. The digital information is essentially a “twin” conclusions.
of the information embedded in the physical system and
can be linked to the physical system throughout its lifecycle
(Grieves and Vickers 2017). The DT is a fully connected
functional virtual representation of the actual manufacturing Background
system and runs in real-time (or near real-time) mode, driven
by data from the relevant system(s). Traditional commissioning and approval of PLC code of an
In this study we developed a hybrid simulation- and data- automated production system occurs before final set-up of
driven DTE and we linked it with operational PLCs through the system by manually debugging and executing online tri-
high-speed Ethernet/IP data communication. This allows for als on the actual system and the associated equipment (i.e.
the injection, collection, aggregation, and analysis of near actuators, sensors, I/Os, cylinders etc.). The documentation
real-time information of input or output (I/O) signals. We of the involved engineering phases is typically conducted
modeled sensor signals and various component reactions independently and engineers from the various disciplines
at a low-abstraction and high-resolution level to generate work sequentially to develop and verify the tasks of the sys-
the operational system behavior. Furthermore, we took tem (Bathelt and Meile 2007; Pellicciari et al. 2009). For
into consideration the physical and logical aspects associ- example, a control and instrumentation engineer that pro-
ated with the components of the real MPS (e.g. structure, grams the PLCs cannot finish the verification and optimiza-
behavior activities, entities, states), as well as the inherent tion of the control code up until the electromechanical engi-
properties of the related elements and devices such as: the neering tasks and functions of the corresponding machinery
MPS station(s) layout, exact positions of sensors and actua- in the plant are complete and deployable (Kim et al. 2009;
tors, part colors and dimensions, kinematics, and geometric Carlsson et al. 2012).
shapes. Developing a control system using a traditional commis-
The major contributions of this paper include (1) a DTE sioning approach, where PLC controllers would connect
for commissioning a mechatronics MPS; (2) a novel hybrid directly to a real world machinery or hardware, can lead to
simulation and data driven modeling approach that combines design and programming conflicts, long set-up times and
Discrete Event Simulation (DES) and Agent Based Mode- additional implementation and modification costs (Auinger
ling (ABM) paradigms for the development and deployment et al. 1999; Schludermann et al. 2000; Versteegt and Ver-
of a DTE that supports design decisions, tests what-if system braeck 2002; Smith and Cho 2008; Makris et al. 2012;
configurations, verifies and validates the actual behavior of Bathelt and Meile 2007; Pellicciari et al. 2009; Carlsson

13
Journal of Intelligent Manufacturing (2021) 32:1899–1911 1901

et al. 2012). Furthermore, traditional commissioning typi- et al. 2010; Smith and Cho 2008; Versteegt and Verbraeck
cally results in delays, poses potential hazardous, and lim- 2002; Auinger et al. 1999).
its testing of the automated production system due to the The main concept behind soft commissioning or HIL,
high level of risks for equipment damages, and could lead is to test the industrial control software by linking the
to unanticipated machinery downtime (Auinger et al. 1999; actual PLCs to the plant simulation or to what is commonly
Schludermann et al. 2000; Versteegt and Verbraeck 2002; referred to as an emulator (Schludermann et al. 2000). This
Smith and Cho 2008; Makris et al. 2012). way it is possible to examine and test the actual behavior
and interactions of the complete automation system with its
Virtual commissioning emulator off-line. The electrical emulations of sensors and
actuators perform as the interface between the emulator and
While technology is constantly evolving and simulation the embedded system under assessment in a closed feed-
modeling of automated production systems is emerging, back loop. Likewise, the system under assessment executes
Virtual Commissioning (VC) methods have developed and its control code by outputting actuator control signals. Any
deployed over the last decade to overcome these challenges. alterations in the control signals will cause changes to vari-
VC is a testing process that utilizes both a virtual plant able values of the emulator.
model and a virtual control system for simulation. Since the The RIL commissioning approach combines an automa-
beginning, the aim of VC is to assist in solving the problems tion system with its physical workplace and simulated PLCs
that evolve when manufacturing systems are integrated with (Auinger et al. 1999). Control logic information about the
PLCs. This process allows the developer to map and repro- automated production system is integrated and synchronized
duce a real-world automated system and its behavior to a vir- within the virtual controller. RIL can be used as a decision
tual world for visualizing, commissioning, and verifying the support tool to diagnose, maintain and debug the automated
PLC code of the system prior to building the physical system production process as well as to explore alternative strategies
(Smith and Cho 2008; Hoffmann et al. 2010). Therefore, to control the automated system (Graebe et al. 1997; Stöck-
the main benefit of using virtual models is the capability to lein et al.2010). Comparable to traditional commissioning,
notice and avoid costly integration problems prior to devel- a major drawback of the RIL approach is that it involves
oping the physical system. Other benefits of VC include time some of the same potential risks as traditional commission-
efficiency, where research shows the potential to save up to ing and can result in additional costs and possible system
75% of the time required compared to traditional commis- downtime since the verification process takes place on the
sioning (Zäh et al. 2006), and it is a cost effective way to actual production system.
execute concurrent and parallel engineering development, Finally, SIL is a completely offline commissioning
testing and debugging in the same virtual plant model and approach in which both the automation production system
control system earlier in the design phase (Schiess 2001). and the PLCs communicate and interact within a virtual
Furthermore, VC allows engineers to perform a wider range environment. Both HIL and SIL methods reduce commis-
of analysis of alternative product concepts, environmental sioning time and human effort (Eskin and Frank 2003),
factors, system configurations, and operational scenarios increase the quality of a production system by reducing
to ensure the system design meets customer requirements. errors in the PLC programs, and minimizing the potential
The ability to perform a broader analysis also allows for damage to equipment and hazards to personnel (Auinger
the effective training of operators, optimization of system et al. 1999; Schludermann et al. 2000; Smith and Cho 2008;
performance, analysis and prediction of system outcomes, Hoffmann et al. 2010).
and generation of error free controlling code, which leads
to a reduction of risk and improved schedule reliability. Digital twins
According to research in the automation industry, compa-
nies can achieve significant economic benefits in addition The term Digital Twin (DT) was introduced to the general
to quality and productivity as soon as they are willing to public for the first time in NASA’s integrated technology
embrace changes towards the implementation of successful roadmap under the Technology Area 11 Modeling, Simula-
VC (Shahim and Møller 2016). To be successful the virtual tion, Information Technology & Processing (Shafto et al.
models must precisely capture the system of interest. There 2010). The concept of the DT is used in a wide range of
are three ways in which a VC can be configured: (1) Soft application domains, such as in industrial production and
Commissioning or Hardware-in-the-Loop (HIL) Commis- management (Uhlemann et al. 2017; Tao et al. 2018; Guo
sioning; (2) Reality-in-the-Loop (RIL) Commissioning; and et al. 2019; Redelinghuys et al. 2019; Tao et al. 2019), in
(3) Complete VC, Software-in-the-Loop (SIL), or Offline supply chain and logistics (Korth et al. 2018; Marmolejo-
Commissioning. Each method requires different approaches Saucedo 2020; Haße et al. 2020), in designing and testing
and technologies to achieve the desired outcomes (Hoffmann vehicles and systems in automotive, aerospace, and defense

13
1902 Journal of Intelligent Manufacturing (2021) 32:1899–1911

industry (Glaessgen and Stargel 2012; Zheng et al. 2020). Methodology of digital twin emulator
The DT concept has also been applied in the design of for virtual commissioning of the festo
smart cities (Farsi et al. 2020), digital museums (Eklund modular production system
et al. 2012), and even for capturing the athletic behavior of
humans (Barricelli et al. 2020). In this section, we describe the methodology that we
Researchers have expressed the need for modeling followed for the development of the DTE for the Festo
requirements and standardized frameworks for generating Modular Production System (MPS) using a novel hybrid
DTs that are specific to a system within a domain (Shao and simulation- and data-driven modeling approach. More
Helu 2020). However, there is a lack of consensus on the specifically, Section “The festo modular production
concept of DT and how to employ it in different domains system (MPS)” provides an overview of the Festo MPS
(Shao et al. 2019). This might be due to its multi-discipli- and the sequence of material flow at three stations. Sec-
nary, multi-domain applicability. The interpretation of the tion “Hybrid simulation modeling approach for the devel-
term DT has continued to evolve as understanding of the opment of a digital twin emulator” describes the DTE
benefits of a DT increases and the tools for use in develop- architecture, including hardware and software specifica-
ing DTs improve (Lim et al. 2019). A literature review that tions, as well as the simulation modeling approaches that
examined 223 papers related to DTs provides an overview of we used in this work. Finally, in Sections “Distribution
various definitions from several use cases that employed DTs station–Sorting station” we describe the models of the dis-
(van der Valk et al. 2020). In addition, Shao et al. (2019) tribution, testing, and sorting stations, respectively.
summarized the views of the simulation community on DTs
for manufacturing concepts, challenges, and different defini-
tions based on different perspectives and examples. The Festo Modular Production System (MPS)
Although, there is no universal agreement on the exact
definition of the term “Digital Twin,” most definitions tend The Festo Modular Production System (MPS) is a robust
to agree on the concept that it is a virtual representation of a training and research platform with significant scaling
physical entity or system. DTs are often viewed as real-time capability. Its modular design allows researchers to break
simulation models that may serve a specific scope. Accord- down complex systems into the most important technolo-
ing to our perspective, Digital Twin is a real-time simulation gies and rapidly design, model, and troubleshoot highly
model of a physical system or a cyber-physical system, asset, complex systems that feature real-world smart factory
product, process, or service connected with (near) real-time technologies. The modular production paradigm enables
data in a closed feedback loop. DTs can assist with captur- concurrent product and production system design and can
ing, understanding, analyzing, improving, optimizing, and further help companies reduce production costs and lead
predicting the behavior of a system, product, process, ser- times (Rogers and Bottaci 1997). Manufacturers that use
vice, or the physical system’s performance. this production concept can combine individual standard-
The DTE described in this paper aims to contribute to ized modular machines in the production process to assem-
a greater effort of the simulation community towards DTs ble new production lines as a response to the increased
of smart manufacturing systems through the utilization of competition and new demands. To ensure an optimal
hybrid simulation approaches for the real-time model gen- design of high productivity in an automated modular pro-
eration and the employment of event-based communication duction line, a well-adjusted sophisticated control program
and synchronization mechanisms for the model synchroni- capable of maintaining speed at high stability and reliabil-
zation with its real manufacturing system counterpart. The ity under perplexed control conditions is required. The use
DTE uses a soft commissioning HIL approach as it is con- of PLC’s allows us to run these complex control programs
nected to the three actual PLCs of the MPS station. The to ensure higher efficiency and reliability of the automated
DTE could also be used for SIL commissioning by linking systems and machinery.
it to a virtual control system and virtual fieldbus emulators For this experiment, we used the Festo MPS with Rock-
(as in (Oppelt and Urbas, 2014)). The authors worked on a well Automation PLCs. The complete MPS is composed
novel way to facilitate a user-friendly approach to the vir- of six stations: distribution, testing, processing, assembly,
tual commissioning process by addressing major challenges sorting, and storing stations as shown in Fig. 1. The sys-
such as, (1) the straightforward and low-cost creation of a tem is configured to produce a workpiece set for cylinder
novel DTE, and (2) its ability to work in tandem with PLC assembly consisting of the body, piston, spring, and cover.
development software. For the purposes of this experiment, we only used the body
of the workpiece and three of the stations: distribution,
testing, and sorting.

13
Journal of Intelligent Manufacturing (2021) 32:1899–1911 1903

Fig. 1  The complete mechatronics modular production station Fig. 2  Sequence of the material flow at the three stations

The process starts by loading ten workpieces with their

open side up in the distribution station magazine. The dis- and silver workpieces to an air slide that transports the
tribution process functions as a feed station with a parting pieces to the sorting station.
device that separates a workpiece from the magazine and The sorting station consists of a conveyor belt and two
makes it available for a subsequent process. The distribu- sorting gates (switches) that actuate to sort workpieces based
tion module consists of a profile plate, a PLC board with on their color, width, and height properties. The parts are
control console, a feed magazine module, and a transfer sorted with red parts placed in the first tray, silver parts in
module. The feed magazine module separates workpieces the second tray, and black parts in the third most distant tray.
from the magazine using a double-action cylinder which Figure 2 illustrates the sequence of the material flow in the
pushes out the bottom workpiece from the gravity feed three stations.
magazine. The end position sensing of the workpiece is
achieved by a magnetic proximity sensor. The fill level Hybrid simulation modeling approach
of the magazine is monitored and four different types of for the development of a digital twin emulator
workpieces that are classified based on different charac-
teristics (height, and color), are supplied, and fed to the Modeling and simulation approaches are often applied to
magazine in any order. The transfer to the test module is enhance decision-making of a factory’s production pro-
executed by a pneumatic handling device. cesses. To provide the most valuable form of decision sup-
When the testing station is available to test a part, a port, a simulation model must accurately represent the sys-
workpiece is picked up by a vacuum suction cup and is tem of interest and be capable of evaluation of a viable set
transferred to the testing station via a swivel drive. The of alternatives. Existing studies, however, focus mostly on
purpose of the testing station is to determine the material the factory layout design (i.e. (Biesinger et al. 2018) and
and color characteristics of a workpiece. The workpiece (Zhang et al. 2019) and consider only simple concepts of
is inspected for color and height. For the material or color the variability of production systems. Additionally, a model
recognition three proximity sensors are used. The first is must contain and assess Key Performance Indicators (KPIs)
an inductive proximity sensor which identifies only metal- of the production process, represent the variability of an
lic workpieces, next a capacitive proximity sensor iden- automation production system, and capture details of its
tifies both the metallic and the plastic workpieces, and variability in a modular and reconfigurable DT (Delbrüg-
finally an optical proximity sensor identifies the red and ger and Rossmann 2019; Schluse et al. 2017). To achieve
the metallic workpieces. Once the logic operations of the the desired results from a simulation model, a new hybrid
output signals determine the respective workpieces, a lift- simulation modelling approach is recommended that can
ing cylinder and an ejecting cylinder lift a workpiece from systematically model the variability of an automated pro-
the material detection module and move it to the measur- duction system and its components in a DTE. Our approach
ing module. The measuring module tests and determines combines DES and ABM to model different variants of the
the height dimensions of a workpiece based on a linear system and the capabilities of its production equipment in a
potentiometer with a voltage divider. The defined analogue modular and reconfigurable DTE. Each valid combination
measured value is processed by the PLC and a slider either of variants results in a 3D-simulable DT for the whole MPS,
rejects the white workpieces that do not meet the height allowing automatic validation testing and fast feedback loops
specifications to the lower tray, or moves the red, black, early in the design process. This hybrid approach is not only

13
1904 Journal of Intelligent Manufacturing (2021) 32:1899–1911

a crucial basis for managing multiple complex system vari- Hybrid DES‑ABM simulation architecture
ants, but also for exploring scenarios and different configu-
rations. This hybrid simulation approach can be a powerful The DTE interacts with DES models, while many HIL sys-
simulation-based optimization approach and cost-efficient tems are based on continuous real-time simulation (Boot
commissioning process of such automated systems. et al.1999). The desire to emulate the physical and logical
aspects of the MPS station resulted in the development of
Hardware and software specifications this innovative hybrid simulation- and -data driven DTE
that combines near real-time data with DES and ABM
The DTE is executed using the Anylogic™ Simulation soft- approaches. For information exchange between the DES
ware installed on a standard gaming laptop without the need and ABM we use value assignment relationships (i.e. ‘direct
of a high-speed Digital Signal Processor that most HIL sys- replacement of values of variables’’) and impact statement
tems require in order to achieve signal response times of less relationships (i.e.‘‘control flow relationships’’) of variables
than 1 ms (Schuette and Waeltermann 2005; Hanselmann that are accurately captured by or influenced by each simula-
1998). The minimum computer requirements for the Any- tion modeling approach (Mykoniatis 2015) (Mykoniatis and
logic™ simulation software are at least 1.5 GB of free space Angelopoulou 2020). This framework served as a guideline
and a modern processor with a minimum of 2 cores, for to establish relationships between interaction points and to
optimal performance. More cores would benefit the parallel collect information about variable definitions and parameters
processing computing of tasks and parallel evaluators that for meeting a variety of VC requirements (i.e. the hybrid
are used in parameterized experiments with multiple simu- model inputs are defined by parameterizing the automation
lation output runs. The AnyLogic multi-method simulation system for the VC process).
software is a Java and Eclipse-based model development Specifically, DES captures the workpiece flow through
environment that can run on Microsoft Windows, Mac, and the production process and the statistical output of the sys-
Linux operating systems. tem. This modeling approach can act as a resource-efficient
During the experimentation period, we used high-speed technique to isolate and detect issues much earlier in the
Ethernet/IP data communication on a local area network design process and sequence of the material flow in the sys-
(LAN) to connect the PLCs and the DTE. Although com- tem. The workpieces are modelled as entities in the DES
munication over the internet is possible, it is not recom- model. Each workpiece is characterized by color type (black,
mended due to potential cyber security issues. The current silver, red, and white), width, and height. We 3D printed the
generation of PLCs has an integrated Ethernet port on the white workpieces with a different height (21 mm) for verify-
controllers (Reszka 2019) which allows remote data collec- ing the rejection process. Figure 3 illustrates the DES model
tion and administration of a control process. This is possible and the animation of the sequence of the material flow in
by utilizing remote I/O on Ethernet-based protocols such as the three stations: distribution, testing, and sorting station.
the Ethernet/IP, Profinet, or Modbus/TCP (UDP) and other The ABM approach is used to model each of the sta-
information technologies such as Web browsers and FTP tions (distribution, testing, sorting) as an agent with various
servers that help us make machines smarter. Advantages of states. In this work, ABM is used as a computational instan-
communication over the internet include remote monitoring, tiation of a complex automation system which exemplifies
diagnostics, and maintenance functions, as well as easier individual agents (stations) and their collective behavior
integration with big data, artificial intelligence, and other (MPS). ABM allows us to focus on the defining attributes
embedded enterprise systems. Internet capability allows for and behavioral rules for each station (i.e., agent) of the MPS
continuous monitoring of operational metrics in a factory in the simulation. In our case, each agent can perform a
and provides convenient access to data and information for list of actions that are linked to their corresponding events
engineers, and management. which control the process flow of the discrete event model.
However, when connecting PLCs to a network with Inter- The state charts are used by the simulation to set inputs in
net access, the devices will be exposed to potential cyber the PLC. The PLC input card(s) are inhibited so that input
security threats. Examples of common security measures variables are not automatically updated. Input variables are
that can help reduce the risk of cyber security threats and configured to read or write so the simulation can write to
maintain the integrity of your system and data may include these inputs. The outputs from the PLC read by the simula-
(Reszka 2019):proper configuration of the router (e.g. keep tion activate the corresponding transitions within each state
all the vulnerable FTP ports closed to prevent uploading a chart.
program that could override the operations of a controller); An agent’s state will move from an active state (light
setup of a virtual private network (VPN) for encrypting the green) into a transition state (light blue) when a PLC
data transmissions over the Internet; and selection of PLCs output is activated. The state transitions are time or
that utilize sophisticated embedded operating systems. condition triggered. The state chart will remain in this

13
Journal of Intelligent Manufacturing (2021) 32:1899–1911 1905

Fig. 3  Discrete event simulation model sequence of the material flow at the three stations

Fig. 5  ABM state-chart of testing station

Fig. 4  ABM state-chart of distribution station

transition state until a time condition transition occurs

which advances the state chart into a new active state.
If the output which has advanced the state chart into the
transition state is deactivated, then the state chart returns
to its original active state. Figures 4, 5, 6 depict the agent-
based state-chart models for each of the three stations.
The ABM state-charts of the DTE use Boolean variables
to transition between states and can modify only Boolean
variables. The Boolean variables can have only two states,
false for inactive and true for active. Analog values can
Fig. 6  ABM state-chart of sorting station

13
1906 Journal of Intelligent Manufacturing (2021) 32:1899–1911

also be modelled; however, they are not required for this Communication and time synchronization
simulation and therefore we did not include them. between the PLCs and the DTE
Distribution station PLC programs can be scheduled as continuous, periodic,
event-based execution, or a mixture of these options. Dur-
The state-chart in Fig. 4 models the behavior of the eject- ing continuous execution, a program runs in an uninter-
ing cylinder position (extended/retracted), the rotary drive rupted manner and, upon completion of a scan, it restarts
position (downstream station position/magazine position), and repeats its logic without delay. Continuous execution
and vacuum (on/off). Initially the ejecting cylinder is in eliminates idle time on the PLC processors and allows
the extended position, the rotary drive is in the “magazine” minimal time for communications.
position and the vacuum is off. The state-chart cycle begins Periodic execution allows a program to run at specified
with the rotary drive swivelling to the “downstream station” intervals set by the programmer. The processor is idle and
position. The ejecting cylinder retracts and ejects a work- can be used for communication between periodic execu-
piece out of the magazine and the rotary drive swivels into tions. The standardized interval between executions per-
“magazine” position. The suction pressure of the vacuum is mits instructions such as proportional–integral–derivative
switched on, the ejecting cylinder extends and releases the (PID) controllers to operate effectively and it allows the
workpiece and the rotary drive swivels back to the down- programmer to configure time critical programs to execute
stream station. Finally, the suction pressure of the vacuum in an appropriate manner. Finally, event-based execution
and the rotary drive return to their initial state or position. prevents a program from executing until an event triggers
execution of the program.
Testing station To synchronize the PLCs and simulation time of the
DTE, periodic and event-based executions are used in
The state-chart in Fig. 5 models the behavior of the test- combination. The PLC code is stored in an event-based
ing station activities which include the lifting cylinder program and is only executed when an event from the syn-
(raised/lowered), the ejecting cylinder (extended/retracted), chronization program triggers it. The synchronization pro-
and the air cushioned slide (on/off). The starting condition gram uses real world time, while the event-based program
of the testing station is with the lifting cylinder lowered, executes in simulation time. To achieve synchronization
the ejecting cylinder retracted, and the air cushioned slide between the PLCs and the DTE, a “PLC_sync” event is
off. The state-chart cycle begins by detecting the available created for each PLC station within the DTE. Figure 7
workpiece in the retainer. Then the color/material of the illustrates the distribution, testing, and sorting PLC Syn-
workpiece (red/metallic) is determined, and the lifting cyl- chronizers with their corresponding mapped I/O variables.
inder is raised. The workpiece height is measured, and the The three “PLC_sync” events execute periodically
air-cushioned slide is switched on. The ejecting cylinder is (every 100 ms) in simulation time and trigger the writ-
extended to push the workpiece through the air-cushioned ing of all input values from the DTE state-charts to the
slide to the sorting station. Then the ejecting cylinder, the PLCs, setting of the PLC system time to the DTE time
air-cushioned slide, and the lifting cylinder return to their using the Set System Value (SSV) instruction, and triggers
initial state/position. the PLC code execution. The scan cycle time depends on
the program content and length. It is important the recur-
Sorting station rence time of the “PLC_sync” events to match the allow-
able cycle time of the PLCs program. When inputs change
The state-chart in Fig. 6 models the behavior of the sort- their state quicker than the scan time, then the PLCs will
ing station activities that include the lifting cylinder (raised/ most likely fail to capture these changes. To avoid latency
lowered), the ejecting cylinder (extended/retracted), and the issues that result from inputs altering asynchronously with
air cushioned slide (on/off). Initially, the stopper is extended, respect to PLC scan cycle, Allen Bradley Control Logix
the conveyor motor is off, and the two branches are retracted. PLCs allow the user to write a short, frequently scanned
The state-chart cycle begins with a defuse sensor which program where this is necessary and to “concurrently”
detects an available workpiece, and the conveyor motor is execute a longer, slower program for non-time-critical
turned on. Then the color/material is identified, the stop- functions.
per is retracted, and the workpiece is ejected. Based on the Once the PLC code is executed, the DTE reads the PLC
identified color, the corresponding branch is extended, and output values from the PLC and updates the DTE out-
the workpiece is deposited on the slide at the end of the put variables used by the corresponding state diagrams,
conveyor. The conveyor motor is turned off, the stopper is described in section distribution station through sorting
extended and branch 1 or branch 2 is retracted.

13
Journal of Intelligent Manufacturing (2021) 32:1899–1911 1907

Fig. 7  PLC Synchronizers with their corresponding mapped I/O variables

station. This way the DTE can communicate in closed reactive actual computations that are carried out by the PLCs
feedback loop with the PLCs, collect data, and execute regarding the real-world physical signals and events.
PLC code in near real time with the MP. The communica- The initiation of the MPS system (Fig. 8) began with a
tion between the MPS and the DTE is achieve through an warm-up period in which the system appeared to achieve
EtherNet IP hardwired connection and EPICS, a set of steady state after 16 workpieces. We set up the period of
open source libraries provided by the Argonne National interest of the processing time of each individual MPS com-
Laboratory (Experimental Physics and Industrial Control ponent after 16 workpieces passed through the system and
Systems 2020). We imported the EPICS set of open source until 400 workpieces were processed. The PLCs stopwatches
libraries to the model dependencies to allow the DTE to automatically collected real-time processing times of each
map the MPS controller tags with the DTE I/O variables MPS component. We analyzed the descriptive statistics of
that are defined within both systems, respectively. The the values and verified against video recording timestamps
ABM state-charts of the DTE use the sensor variables to of the corresponding processes and regular reviews of the
identify components of the MPS state and capture the tran- manufacturer’s documented specifications. We successfully
sition between system states and the actuators for affecting tested the responses of the DTE to verify the correct work-
control. Consequently, PLC tags and DTE variables are flow at each station, as well as the total time to process work-
essential to enable the exchange of information between pieces in system for each type of workpiece. Furthermore,
the PLC controllers and the DTE as well as for developing we viewed the animation of the simulation output (Fig. 9),
distributed soft real-time control systems and collecting the state diagrams and their corresponding I/O state dia-
real-time control data. gram variables, to visually inspect and verify the DTE. We

Experimentation and results

The literature discussed several existing formal methods and

frameworks to verify and validate the behaviors of auto-
mated production systems during development. (Hanisch
et al. 2006; Lampérière-Couffin et al. 1999; Younis and Frey
2003; Dahl et al. 2016). In our approach we apply a regres-
sion verification technique that requires less functional spec-
ifications and reduces the complexity of the verification and
validation of the DTE model. The high fidelity of the DTE
depends on the inner validity of the model, which is how
well it matches the data collected and how well the collected
data capture the details of each individual MPS component
and its interactions on which the model sensitivity depends.
Therefore, near real-time data collection is critical to the
effectiveness of the simulation-based emulator to match the Fig. 8  Festo modular production system

13
1908 Journal of Intelligent Manufacturing (2021) 32:1899–1911

that the difference in population means is between − 0.541

and 0.265 s for the metallic workpiece, − 0.586 and 0.010 s
for the red workpiece, and − 0.3402 and 0.0402 s for the
black workpiece. To determine whether these differences
between the population means are statistically significant,
we compared the p-values to the significance level of 0.05.
The p-values for the metallic, red, and black workpiece are
0.467, 0.057, and 0.111, respectively. In all three cases, the
p-values are greater than the level of significance, and there-
fore we cannot reject the null hypothesis that the difference
between the means is zero. The confidence intervals of the
Fig. 9  Digital twin emulator of the modular production station during times overlap and no statistical difference between actual
3D animation MPS records and DTE results is observed. Consequently,
the DTE was considered valid. Validating the DTE allows
for the development of alternative scenarios that can provide
also examined the quantitative Key Performance Indicators sufficient evidence to show that implementing them in the
(KPIs), such as throughput, equipment usage, and capaci- physical system may improve or deteriorate the system.
ties for validity. We computed the sample mean and sample
variance for each KPI of the DTE and compared with the
actual observed MPS mean and variance (Tables 1, 2, 3).
Therefore, we performed two-sample t-tests, to compare the
Conclusions and future work
two averages (means) for each colored workpiece. The null
This paper described the design, development, and imple-
hypothesis was that the difference between the two means
mentation of a DTE of an automated mechatronic MPS that
is zero (Η0: μ1 − μ2 = 0), and the alternative hypothesis that
is linked with PLCs and allows for the exchange of near real-
the difference between the two means is not equal to zero
time information with the physical system. The development
(H1: μ1-μ2 ≠ 0). Then, we estimated the difference in the
and deployment of the DTE involved a novel hybrid simu-
population means using the confidence interval for each
lation- and data-driven modeling approach that combines
colored workpiece. The estimate of the difference in popu-
DES and ABM paradigms. The DTE can support design
lation means for the total time in the system is 0.138 s for the
decisions, test what-if system configurations, verify and
metallic workpiece, 0.288 s for the red workpiece, and 0.15 s
validate the actual behavior of the complete system off-line,
for the black workpiece. Thus, we can be 95% confident

Table 1  Comparison of real world and DTE Perfrmance Measures for Metallic workpiece
Measure name MPS average MPS standard devia- MPS confidence inter- DTE average DTE standard devia- DTE C.I.
tion val (C.I.)) tion

Average total time in 9.756 0.663 [9.637, 10.039] 9.894 0.15 [9.773, 10.015]
system

Table 2  Comparison of real world and DTE Perfrmance Measures for Red workpiece

Measure name MPS average MPS standard devia- MPS confidence inter- DTE average DTE standard devia- DTE C.I.
tion val (C.I.)) tion

Average total time in 9.119 0.456 [9.296, 9.662] 9.407 0.474 [8.933, 9.590]
system

Table 3  Comparison of real world and DTE Perfrmance Measures for black workpiece
Measure name MPS average MPS standard deviation MPS confidence interval DTE average DTE standard deviation DTE C.I.
(C.I.))

Average total time 10.072 0.296 [9.919,10.225] 10.222 0.15 [10.106, 0.338]

13
Journal of Intelligent Manufacturing (2021) 32:1899–1911 1909

test realistic reactions and provide statistics of system’s per- References

formance metrics.
During the early design and development phases of an Andersen, A.-L., Nielsen, K., & Brunoe, T. D. (2016). Prerequisites
automated production system, many industries remain in and barriers for the development of reconfigurable manufactur-
ing systems for high speed ramp-up. Procedia Cirp, 51(1), 7–12.
a relatively primitive stage, using costly tools and time- Anglani, A., Grieco, A., Pacella, M., & Tolio, T. (2002). Object-ori-
consuming techniques for testing and verifying their con- ented modeling and simulation of flexible manufacturing systems:
trol systems and code. Furthermore, many businesses and a rule-based procedure. Simulation Modelling Practice and The-
organizations still use inadequate historical data to predict ory, 10(3–4), 209–234.
Auinger, F., Vorderwinkler, M., & Buchtela, G. (1999). Interface driven
KPI’s relevant to their automated production systems. Con- domain-independent modeling architecture for “soft-commission-
sidering the increased dynamic complexity and features that ing” and “reality in the loop”. WSC’99. Winter Simulation Confer-
will be carried out in the future of mechatronics systems, ence Proceedings “Simulation-A Bridge to the Future” (Cat. No.
commission of the control systems as well as prediction of 99CH37038). 1. IEEE.
Barricelli, B. R., Casiraghi, E., Gliozzo, J., Petrini, A., & Valtolina,
their KPIs can only be obtained by limited computation or S. (2020). Human Digital Twin for Fitness Management. IEEE
empirical ways. The traditional approach cannot make use Access, 8, 26637–26664.
of all historical production data and, as a result, may cause Bathelt, J., & Meile, J. (2007). Computer aided methods supporting
integration issues and design flaws that can lead to consider- concurrent engineering when designing mechatronic systems con-
trolled by a plc. In Proceedings of International Conference on
able problems in future manufacturing. Due to our hybrid Manufacturing Automation (ICMA07). Singapore.
modular DTE approach many inputs can be parameterized. Beckert, B., Mund, J., Ulbrich, M., & Weigl, A. (2019). Formal Verifi-
Outputs such as product throughput, workpiece cycle time, cation of Evolutionary Changes. In Managed Software Evolution
equipment utilization, etc. can be obtained depending on dif- (pp. 310–311). Springer: Cham.
Biesinger, F., Meike, D., Kraß, B., & Weyrich, M. (2018). A Case
ferent inputs combinations. This hybrid DES-ABM approach Study for a Digital Twin of Body-in-White Production Systems
can help overcome challenges and limitations that stand- General Concept for Automated Updating of Planning Projects in
alone simulation languages often face in capturing the con- the Digital Factory. 2018 IEEE 23rd International Conference on
trol logic of an automated production system (Anglani et al. Emerging Technologies and Factory Automation (ETFA). 1, (pp.
19–26). Turin: IEEE.
2002). This hybrid simulation modeling approach facilitates Boot, R., Richert, J., Schutte, H., & Rukgauer, A. (1999). Automated
DT as a strong tool for insight into the design of future pro- test of ECUs in a hardware-in-the-loop simulation environment.
duction systems and supports the designers for creation of In Proceedings of the 1999 IEEE International Symposium on
optimized designs. Therefore, a DTE such as the one pre- Computer Aided Control System Design (Cat. No. 99TH8404),
(pp. 587–594). IEEE.
sented here is a reasonable and proper solution due to its Carlsson, H., Svensson, B., Danielsson, F., & Lennartson, B. (2012).
high fidelity to the physical system (Tao et al. 2018). Manu- Methods for reliable simulation-based PLC code verification.
facturers can prevent substantial installed systems problems IEEE Transactions on Industrial Informatics, 8(2), 267–278.
earlier in the design and development phases by implement- Dahl, M., Bengtsson, K., Bergagård, P., Fabian, M., & Falkman, P.
(2016). Integrated virtual preparation and commissioning: sup-
ing a DTE and revealing the resulting actual performance porting formal methods during automation systems development.
the system can achieve. DTEs can undeniably enhance the IFAC-Papers OnLine, 49(12), 1939–1944.
viability and quality of a production system as it forces the Delbrügger, T., & Rossmann, J. (2019). Representing adaptation
engineers and programmers to design and assess the system options in experimentable digital twins of production systems.
International Journal of Computer Integrated Manufacturing,
in an interdisciplinary and aggregate manner and its control 32(4–5), 352–365.
code in detail. A DTE supports the coordination and integra- Eklund, P., Wray, T., Goodall, P., & Lawson, A. (2012). Design, infor-
tion of automatic system components and makes the effort of mation organisation and the evaluation of the Virtual Museum of
developing, verifying, and validating an automated manufac- the Pacific digital ecosystem. Journal of Ambient Intelligence and
Humanized Computing, 3(4), 265–280.
turing system consistent, comprehensible, and maintainable. Eskin, A., & Frank, A. (2003). Offline simulation of a managed system
In future work, we will include testing of what-if system for testing a developed management system. In Proceedings 2003
configurations to explore potential limitations of the com- Symposium on Security and Privacy, (pp. 154–163). IEEE.
munication and synchronization approach that we adopted Experimental Physics and Industrial Control Systems. (2020).
(Argonne National Laboratory) Retrieved April 2019, from https​
in this work. Specifically, we plan to add processing, buffer, ://epics​.anl.gov
and robotic assembly stations between the testing and the Farsi, M., Daneshkhah, A., Hosseinian-Far, A., & Jahankhani, H.
sorting stations to understand, capture, and analyze more (Eds.). (2020). Digital Twin Technologies and Smart Cities. Ber-
complex systems and behaviors. Additionally, the applica- lin: Springer.
Glaessgen, E., & Stargel, D. (2012). The digital twin paradigm for
tion of this approach in the development of a complex sys- future NASA and US Air Force vehicles. 53rd AIAA/ASME/ASCE/
tem in different domains is of significant interest. AHS/ASC structures, structural dynamics and materials confer-
ence 20th AIAA/ASME/AHS adaptive structures conference 14th,
Acknowledgements The authors would like to thank Mr Richard Sack (p. 1818).
for his helpful suggestions and programming efforts.

13
1910 Journal of Intelligent Manufacturing (2021) 32:1899–1911

Graebe, S., Schleicher, M., Steringer, R., & Zeichen, G. (1997). Reality International Journal on Interactive Design and Manufacturing
in the Loop. Manufacturing System Design. (IJIDeM), 3(2), 81–91.
Grieves, M., & Vickers, J. (2017). Digital twin: Mitigating unpredict- Redelinghuys, A. J. H., Basson, A. H., & Kruger, K. (2019). A six-
able, undesirable emergent behavior in complex systems. Trans- layer architecture for the digital twin: a manufacturing case
disciplinary perspectives on complex systems (pp. 85–113). Cham: study implementation. Journal of Intelligent Manufacturing,
Springer. 31, 1383–1402. https​://doi.org/10.1007/s1084​5-019-01516​-6
Guo, J., Zhao, N., Sun, L., & Zhang, S. (2019). Modular based flexible Reszka, P. (2019, November 21). Automation.com. (WAGO Cor-
digital twin for factory design. Journal of Ambient Intelligence poration) Retrieved 8 4, 2019, from https​://www.autom​ation​
and Humanized Computing, 10(3), 1189–1200. .com/en-us/artic​les/2019/conne​cting​-to-the-plc-from-a-remot​
Hanisch, H.-M., Lobov, A., Lastra, J. L., Tuokko, R., & Vyatkin, V. e-locat​ion
(2006). Formal validation of intelligent-automated production Rogers, G. G., & Bottaci, L. (1997). Modular production systems: a
systems: towards industrial applications. International Journal new manufacturing paradigm. Journal of Intelligent Manufactur-
of Manufacturing Technology and Management, 1(3), 75–106. ing, 8(2), 147–156.
Hanselmann, H. (1998). Hardware-in-the-loop simulation testing and Schiess, C. (2001). Emulation: debug it in the lab-not on the floor. In
its integration into a CACSD toolset. In Proceedings of Joint Con- Proceeding of the 2001 Winter Simulation Conference (Cat. No.
ference on Control Applications Intelligent Control and Computer 01CH37304). 2. IEEE.
Aided Control System Design (pp. 152–156). IEEE. Schludermann, H., Kirchmair, T., & Vorderwinkler, M. (2000). Soft-
Haße, H., Li, B., Weißenberg, N., Cirullies, J., & Otto, B. (2020). commissioning: hardware-in-the-loop-based verification of con-
Digital twin for real-time data processing in logistics. In Artificial troller software. In Proceedings of the 32nd conference on Win-
Intelligence and Digital Transformation in Supply Chain Manage- ter simulation. (pp. 893–899). Society for Computer Simulation
ment: Innovative Approaches for Supply Chains. Proceedings of International.
the Hamburg International Conference of Logistics (HICL). 27, Schluse, M., Atorf, L., & Rossmann, J. (2017). Experimentable digital
(pp. 4–28). Berlin: epubli GmbH. twins for model-based systems engineering and simulation-based
Hoffmann, P., Schumann, R., Maksoud, T. M., & Premier, G. C. (2010). development. In 2017 Annual IEEE International Systems Confer-
Virtual commissioning of manufacturing systems a review and ence (SysCon) (pp. 1-8). IEEE.
new approaches for simplification. In European Conference of Schuette, H., & Waeltermann, P. (2005). Hardware-in-the-loop testing
Modeling and Simulation (ECMS), (pp. 175–181). of vehicle dynamics controllers–a technical survey. SAE Transac-
Kim, H.-J. S., Kernbaum, & Seliger, G. (2009). Emulation-based con- tions, 2, 593–609. https​://doi.org/10.4271/2005-01-1660
trol of a disassembly system for LCD monitors. The International Shafto, M., Conroy, M., Doyle, R., Glaesgen., Ed; Kemp, C., LeM-
Journal of Advanced Manufacturing Technology, 3(4), 383–392. oigne, J., Wang, L. (2010, November). DRAFT Modeling, Simula-
Korth, B., Schwede, C., & Zajac, M. (2018). 2018 Simulation-ready tion, Information Technology & Processing Roadmap. National
digital twin for realtime management of logistics systems. In Aeronautics and Space Administration. Retrieved 8 10, 2020,
IEEE International Conference on Big Data (Big Data) (pp. from nasa.gov: https​://www.nasa.gov/pdf/50132​1main​_TA11-
4194–4201). IEEE. MSITP​-DRAFT​-Nov20​10-A1.pdf
Lampérière-Couffin, S., Rossi, O., Roussel, J.-M., & Lesage, J.-J. Shahim, N., & Møller, C. (2016). Economic justification of virtual
(1999). Formal validation of PLC programs: a survey. In 1999 commissioning in automation industry. In 2016 Winter Simulation
European Control Conference (ECC) (pp. 2170–2175). IEEE. Conference (WSC) (pp. 2430–2441). IEEE.
Lee, C. G., & Park, S. C. (2014). Survey on the virtual commissioning Shao, G., & Helu, M. (2020). Framework for a digital twin in manu-
of manufacturing systems. Journal of Computational Design and facturing: scope and requirements. Manufacturing Letters, 24,
Engineering, 1(3), 213–222. 105–107.
Lim, K. Y. H., Zheng, P., & Chen, C. H. (2019). A state-of-the-art Shao, G., Jain, S., Laroque, C., Lee, L. H., Lendermann, P., & Oliver,
survey of Digital Twin: techniques, engineering product life- R. (2019). Digital twin for smart manufacturing: the simulation
cycle management and business innovation perspectives. Jour- aspect. 2019 Winter Simulation Conference (WSC) (pp. 2085–
nal of Intelligent Manufacturing, 31, 1313–1337. https​://doi. 2098). IEEE.
org/10.1007/s1084​5-019-01512​-w Smith, J. S., & Cho, Y. (2008). Offline commissioning of a PLC-based
Makris, S., Michalos, G., & Chryssolouris, G. (2012). Virtual control system using Arena. In Winter Simulation Conference.
commissioning of an assembly cell with cooperating robots. IEEE.
Advances in Decision Sciences, 2012, 428060. https ​ : //doi. Stöcklein, J., Geiger, C., & Paelke, V. (2010). Mixed reality in the
org/10.1155/2012/42806​0 loop—design process for interactive mechatronical systems. IEEE
Marmolejo-Saucedo, J. A. (2020). Design and Development of Digi- Virtual Reality Conference (VR). IEEE.
tal Twins: a Case Study in Supply Chains. Mobile Networks and Tao, F., Cheng, J., Qi, Q., Zhang, M., Zhang, H., & Sui, F. (2018). Digi-
Applications. 1. Springer tal twin-driven product design, manufacturing and service with
Mykoniatis, K. (2015). A Generic Framework For Multi-Method Mod- big data. The International Journal of Advanced Manufacturing
eling and Simulation of Complex Systems Using Discrete Event, Technology, 94(9–12), 3563–3576.
System Dynamics and Agent Based Approaches. Tao, F., Sui, F., Liu, A., Qi, Q., Zhang, M., Song, B., et al. (2019). Digi-
Mykoniatis, K., & Angelopoulou, A. (2020). A modeling framework tal twin-driven product design framework. International Journal
for the application of multi-paradigm simulation methods. Simula- of Production Research, 57(12), 3935–3953.
tion, 96(1), 55–73. Thramboulidis, K. (2005). Model-integrated mechatronics-toward
Oppelt, M., & Urbas, L. (2014). Integrated virtual commissioning an a new paradigm in the development of manufacturing systems.
essential activity in the automation engineering process: From IEEE Transactions on Industrial Informatics, 1(1), 54–61.
virtual commissioning to simulation supported engineering. In Uhlemann, T. H.-J., Christian, L., & Rolf, S. (2017). The digital twin:
IECON 2014–40th Annual Conference of the IEEE Industrial Realizing the cyber-physical production system for industry 4.0.
Electronics Society (pp. 2564–2570). IEEE. Procedia Cirp 61, (pp. 335–340).
Pellicciari, M., Andrisano, A. O., Leali, F., & Vergnano, A. (2009). van der Valk, H., Hendrik, H., Möller, F., Arbter, M., Henning, J.-L.,
Engineering method for adaptive manufacturing systems design. & Otto, B. (2020). A Taxonomy of Digital Twins. In 26th Ameri-
cas Conference on Information Systems (AMCIS) At: Salt Lake

13
Journal of Intelligent Manufacturing (2021) 32:1899–1911 1911

City, USA, Projects: Leistungszentrum Logistik und IT Reference Zhang, Z., Wang, X., Wang, X., Cui, F., & Hui, C. (2019). A simula-
Architecture for Digital Twins in Logistics Lab: Boris Otto’s Lab. tion-based approach for plant layout design and production plan-
https​://www.resea​rchga​te.net/publi​catio​n/34123​5159_A_Taxon​ ning. Journal of Ambient Intelligence and Humanized Computing,
omy_of_Digit​al_Twins​ 10(3), 12171230.
Versteegt, C., & Verbraeck., A. (2002). The extended use of simulation Zheng, Y., Chen, L., Lu, X., Sen, Y., & Cheng, H. (2020). Digital twin
in evaluating real-time control systems of AGVs and automated for geometric feature online inspection system of car body-in-
material handling systems. In Proceedings of the Winter Simula- white. International Journal of Computer Integrated Manufactur-
tion Conference. 2, pp. 1659–1666. IEEE. ing. https​://doi.org/10.1080/09511​92X.2020.17366​37
Younis, M. B., & Frey, G. (2003). Formalization of existing PLC pro-
grams: A survey. In Proceedings of CESA, (pp. 0234–0239). Publisher’s Note Springer Nature remains neutral with regard to
Zäh, M. F., Wünsch, G., Hensel, T., & Lindworsky, A. (2006). Ben- jurisdictional claims in published maps and institutional affiliations.
efits of the Virtual Commissioning–An Experiment (original title:
Nutzen der virtuellen Inbetriebnahme: Ein Experiment). ZWF
Zeitschrift für wirtschaftlichen Fabrikbetrieb, 101(9), 595–599.

13