Blockchain-based Digital Twins. Requirements and Design Decisions for the Data Storage

Masterarbeit , 2022 , 128 Seiten

Informatik - Sonstiges

The following research goal is the development of design decisions for the blockchain-based storage of digital twin data. The digital twin can be deployed in almost every industry and is able to represent any object, being, process or organisation. The resulting benefits are versatile and can encompass a higher transparency and efficiency, cost reduction, or the enabling of completely new functions such as the virtual commissioning. Digital twins do even contribute to a sustainable industrial production as they enable a corresponding reduction of the primary energy emissions of up to eight percent. Due to this substantial added value, digital twins pertain to the top ten strategic technology trends and are expected to reach a market size of over 48 billion USD by 2026.

The existence of digital twins is strongly dependent on its underlying data, for which reason the data storage is of fundamental importance for the operation of this technology. In the past, the digital twin concept mostly relied upon a traditional storage infrastructure in form of databases or clouds. However, digital twins and its data storage face major challenges, e.g., through a
more and more connected world and therefore an increasing number of potential participants and data volume, or as a consequence of growing data security-related risks. Furthermore, potential digital twin use cases entail varying conditions, for which reason alternative storage solutions must be assessed.

The blockchain technology might constitute a possible alternative to realize the storage of digital twin data. Through the novel and nearly unique characteristics such as immutability, decentralisation, or manipulation safety, blockchains could introduce new opportunities to address digital twin use cases. Those characteristics seem relevant at first sight as the application areas
of digital twins become more manifold– especially as the digital twin concept evolves from a descriptive to a rather actionable approach and include more critical data. How the blockchain technology can actually deal with the requirements for the data storage including the abovementioned challenges must therefore be evaluated in order to facilitate
their interplay and potentially achieve their full potential. Hereby, at first the data storage in digital twins must be understood comprehensively to assess the suitability of their deployment.

Table of Contents

LIST OF F IGURES

LIST OF T ABLES

LIST OF A BBREVIATIONS

1 INTRODUCTION
1.1 Background
1.2 Problem Statement
1.3 Scope and Delimitation
1.4 Research Method
1.5 Structure of the Thesis
2 T HEORETICAL B ACKGROUND
2.1 Digital Twin
2.1.1 Origin, Definition, and Characteristics of Digital Twins
2.1.2 Underlying Technologies
2.1.3 Data Storage in Digital Twins
2.1.4 Benefits and Application Areas
2.2 Blockchain
2.2.1 Foundations of the Blockchain Technology
2.2.2 Different Blockchain Configurations
2.3 Blockchain-based Digital Twins
2.3.1 Rationale for Combining Blockchains and Digital Twins
2.3.2 Benefits
2.3.3 State of the Art

3 METHODOLOGY
3.1 Systematic Literature Review and Classification
3.2 Elaboration of Requirements for the Digital Twin Data Storage
3.2.1 Review of the Distribution of Data-related Properties
3.2.2 Expert Interviews
3.3 Evaluation of the Suitability of Blockchain-based Digital Twins
3.4 Development of Design Decisions

4 RESEARCH R ESULTS
4.1 Taxonomy for the Classification of Data-related Properties in Digital Twins
4.1.1 Scope
4.1.2 Data Properties
4.1.3 Data Collection and Communication
4.1.4 Storage and Usage
4.2 Requirements for the Data Storage in Digital Twins
4.2.1 Requirements Derived from Classified Data-related Properties
4.2.2 Requirements Derived from Industry Expert Interviews

5 DISCUSSION
5.1 RQ1: Classification of Data-related Properties in Digital Twins
5.2 RQ2: Requirements for the Data Storage in Digital Twins
5.3 RQ3: Suitability of the Blockchain Technology for Storing Digital Twin Data
5.4 RG: Design Decisions for the Blockchain-based Storage of Digital Twin Data
5.5 Critical Reflexion of the Overall Approach

6 CONCLUSION
6.1 Summary
6.2 Limitations
6.3 Outlook

BIBLIOGRAPHY

LIST OF A PPENDICES

List of Figures

Figure 1: Digital Twin concept by Michael Grieves (Figure sourced from Grieves and Vickers 2017, p. 93)

Figure 2: Differentiation of digital model, digital shadow, and digital twin according to Uhlenkamp et al. (2019) and Kritzinger et al. (2018)

Figure 3: High-level processes in digital twins by VanDerHorn and Mahadevan (2021, p. 3)

Figure 4: Classification of blockchains inspired by Hileman and Rauchs (2017, p. 23)

Figure 5: Example of a blockchain structure (Zheng et al. 2018, p. 355)

Figure 6: Working principle of a blockchain (Prinz et al. 2018, p. 313)

Figure 7: Research trends of blockchain-based digital twins in industry according to Suhail et al. (2021b, p. 8)

Figure 8: Literature search process including filtering

Figure 9: Paper distribution by year of publication

Figure 10: Taxonomy development method by Nickerson et al. (2013, p. 10)

Figure 11: Procedure for the qualitative content analysis by Mayring (2015) and Pfeiffer (2018)

Figure 12: Structure of digital twin data according to Al-Ali et al. (2020)

Figure 13: Illustration of the hierarchic structure of the West Cambridge Site Digital Twin by Lu et al. (2020)

List of Tables

Table 1: Overview of final interview partners

Table 2: Taxonomy part 1 (Scope)

Table 3: Taxonomy part 2 (Data properties)

Table 4: Taxonomy part 3 (Data collection and communication)

Table 5: Taxonomy part 4 (Storage and usage)

Table 6: Overview of requirements for the data storage in digital twins

List of Abbreviations

Abbildung in dieser Leseprobe nicht enthalten

1 Introduction

1.1 Background

“The Digital Twin is the key technology of Industry 4.0” (IDTA 2022). This statement reflects the current and future relevance of digital twins only to some extent. In fact, the digital twin can be deployed in almost every industry and is able to represent any object, being, process or organisation. The resulting benefits are versatile and can encompass a higher transparency and efficiency, cost reduction, or the enabling of completely new functions such as the virtual com­missioning. Digital twins do even contribute to a sustainable industrial production as they ena­ble a corresponding reduction of the primary energy emissions of up to eight percent (Bitkom and Accenture 2021). Due to this substantial added value, digital twins pertain to the top ten strategic technology trends according to Gartner (2019) and are expected to reach a market size of over 48 billion USD by 2026 (MarketsandMarkets 2020).

The existence of digital twins is strongly dependent on its underlying data, for which reason the data storage is of fundamental importance for the operation of this technology. In the past, the digital twin concept mostly relied upon a traditional storage infrastructure in form of databases or clouds. However, digital twins and its data storage face major challenges, e.g., through a more and more connected world and therefore an increasing number of potential participants and data volume, or as a consequence of growing data security-related risks. Furthermore, po­tential digital twin use cases entail varying conditions, for which reason alternative storage so­lutions must be assessed.

The blockchain technology might constitute a possible alternative to realize the storage of dig­ital twin data. Through the novel and nearly unique characteristics such as immutability, decen­tralisation, or manipulation safety, blockchains could introduce new opportunities to address digital twin use cases. Those characteristics seem relevant at first sight as the application areas of digital twins become more manifold- especially as the digital twin concept evolves from a descriptive to a rather actionable approach and include more critical data (Grieves and Vickers 2017). How the blockchain technology can actually deal with the requirements for the data storage including the abovementioned challenges must therefore be evaluated in order to facil­itate their interplay and potentially achieve their full potential. Hereby, at first the data storage in digital twins must be understood comprehensively to assess the suitability of their deploy­ment.

1.2 Problem Statement

Both the blockchain technology and the digital twin concept are still in its infancy in terms of its scientific and industrial coverage. The two technologies are individually deployed in indus­trial applications and generate an added value for the companies but do only rudimentarily de­velop their full potential. Even though the data storage is a fundamental requirement for the existence and functionality of digital twins, research about its realization by means of the block­chain technology is strongly underrepresented. Furthermore, to the best of my knowledge, there is no marketable industry project that exploits the interplay of both technologies.

The academic literature contains a handful approaches to classify digital twins (e.g., van der Valk et al. 2020; Uhlenkamp et al. 2019; Kritzinger et al. 2018), however, mainly with a generic focus (type of paper, purpose, technology etc.) instead of a data-oriented alignment. Similarly, favourable conditions for a general database are present in the literature but digital twin-specific requirements for the data storage are missing. Furthermore, a discussion about the suitability of the blockchain technology for the storage of digital twin data was only rudimentarily led in the literature. Finally, there exist academic literature that cover the influence of blockchain design decisions on its quality attributes (e.g., Xu et al. 2019) or explore the current blockchain-based digital twin approaches (Suhail et al. 2021b). Nonetheless, they lack a discussion about block­chain design decisions depending on the form of digital twin application.

The abovementioned research gap hampers a comprehensive understanding of the data storage in digital twins and the successful adoption of blockchain-based digital twins. As a result, the focal point of this work addresses the data storage of digital twins as well as the intersection of the blockchain technology and the digital twin concept. The aim is to extend the knowledge base regarding the two technologies and foster the viability of an industrial implementation of blockchain-based digital twins. Therefore, the following research goal is targeted in this work:

Development of design decisions for the blockchain-based storage of digital twin data In order to address this research goal properly, the following tantamount research question are answered beforehand:

(1) How can data-related properties in digital twins be classified?
(2) What are requirements for the data storage in digital twins?
(3) Is the blockchain technology suitable for the data storage in digital twins?

1.3 Scope and Delimitation

As already mentioned, the focus in this work is not placed on the digital twin in general but strongly on its data storage. This emphasis is further reflected in the classification of the digital twins, as only data-related properties are worked out. Furthermore, the basis for the classifica­tion comprises solely scientific publications because business-related publications about indus­trial projects are completely missing.

The subsequent main topics, i.e., the requirements for the data storage and the discussion about the suitability and design decisions for blockchain-based digital twins, follow an approach that addresses both the scientific and industrial area. Here, the results and discussions are oriented towards non-human digital twins. This is justified by the higher importance for practitioners and the expected differences in terms of the legal requirements and short-term feasibility. How­ever, particular results such as the findings regarding sensitive data can be transferred.

The evaluation phase of a classical design-oriented information systems research approach in form of a real-world realization or experiment is not covered in the context of this work. Thus, the development of design decisions is based on a high-level discussion, for which reason an in-depth analysis of implementation details is not within the scope of this work. Further infor­mation about the methodology is presented in the following chapter.

1.4 Research Method

In this work, a mixed-method approach was applied to target the different research issues. First, a behaviourist approach provides the foundation for the first two research questions. Data-re- lated properties in digital twins were classified by means of a systematic literature review of scientific publications about digital twin approaches and a subsequent taxonomy development. The requirements for the data storage were elaborated through interviews with digital twin in­dustry experts and a subsequent qualitative content analysis. The extracted requirements were enriched by findings that have been derived by the previous classification of data-related prop­erties.

The methodology for the answer to the last research question and research goal was stronger aligned to the design-oriented information systems research. Hereby, the findings from the first two research issues were incorporated in order to facilitate the subsequent discussion. First, the suitability of a blockchain-based storage of digital twin data was discussed by comparing the requirements for the digital twin data storage with the features, benefits, and drawbacks that are associated with the blockchain technology. Finally, for the answer of the research goal, relevant data-related properties of digital twins and different blockchain variations were juxtaposed. In other words, the most appropriate blockchain designs (concerning the blockchain platform, consensus protocol etc.) for different data-related properties of digital twins were discussed based on the impact on the corresponding quality attributes.

1.5 Structure of the Thesis

Besides the introduction, this work is organized into five distinct sections. First, a theoretical foundation regarding the digital twin concept, the blockchain technology and their combination is provided in order to establish a common understanding of the terminology (Chapter 2). Sub­sequently, chapter 3 details the four components of the methodology that address each research area accordingly. Chapter 4 is dedicated to the elaborated research results and is subdivided into two broad parts. On the one hand, the classification of data-related properties of digital twins is presented in form of a taxonomy. On the other hand, an overview about the identified requirements for the data storage in digital twins is provided. The critical reflexion of the results and methodology follows in chapter 5. In addition, the suitability of blockchain-based digital twins as well as related design decisions are discussed and reflected upon in this section. The thesis is completed by a conclusion, which contains a summary and limitations of the findings as well as an outlook (Chapter 6). Hereby, the objective is to foster an adequate usability of the results and provide suggestions for future research endeavours.

2 Theoretical Background

This chapter aims at building a theoretical foundation for a broad understanding of the discussed topics. First, the concept of the digital twin is explored, and underlying technologies are pre­sented. Subsequently, different forms and challenges of the data storage in digital twins as well as emerging benefits and application areas in general are outlined. Second, the blockchain tech­nology is introduced and related design options are presented. Last, the interconnection of both technologies is illustrated in terms of its relevance respectively potential benefits and a short overview about the state of the art of blockchain-based digital twins is presented.

2.1 Digital Twin

2.1.1 Origin, Definition, and Characteristics of Digital Twins Origin

The origin of the digital twin concept stems from NASA's Apollo program. Concretely, they built at least two congruent space vehicles in order to allow “the engineers to mirror the condi­tions of the space vehicle during the mission” (Boschert and Rosen 2016, p. 63). The vehicle, which remained on earth, was called “twin”. The purpose of the twin comprised two elements. First, the twin was used for the training of flight operations. Second and in this context more interesting, NASA used the vehicle on earth to simulate alternatives based on the space vehicle flight data with the aim of deriving recommendations of action and therefore assisting the as­tronauts in the orbit (Boschert and Rosen 2016). Hereby, the earth vehicle did not impersonate a digital but a physical representation and therefore lacked a direct connection between both objects.

However, it is commonly accepted that the digital twin concept was first informally established during a product lifecycle management presentation from Michael Grieves at the University of Michigan in 2002, who denoted the concept later as “digital equivalent to a physical product”. The concept, as depicted in Figure 1, possesses all main components compared to a modern digital twin (Grieves and Vickers 2017; Liu et al. 2021b). This includes “real space, virtual space, the link for data flow from real space to virtual space, the link for information flow from virtual space to real space and virtual sub-spaces” (Grieves and Vickers 2017, p. 93).

Abbildung in dieser Leseprobe nicht enthalten

Figure 1: Digital Twin concept by Michael Grieves (Figure sourced from Grieves and Vickers 2017, p. 93) Definition

Nowadays, the scientific literature lacks a consistent terminology for the digital twin because of its varying definitions and related concepts such as the Digital Model or Digital Shadow, for which reason an accurate differentiation of the digital twin concept is impeded (Wagner et al. 2017). Hence, to provide a consistent understanding of the digital twin term, the following gen­eralized definition, which was first introduced by Glaessgen and Stargel (2012), is applied in this work:

“A digital twin is an integrated multiphysics, multiscale probabilistic ultra-realistic simulation of systems or products which can mirror the life of its corresponding twin using available physical models, history data, real time data, etc.” (Tao et al. 2019, p. 3938)

However, in this context the term system respectively product is construed in a wide frame, i.e., it includes processes, humans, or organisations as well (Theis et al. 2021). The previously men­tioned notions digital model, digital shadow and digital twin can be further distinguished as follows: The term digital model occurs when both communication links are manual. Hence, there is no automatic data communication between both objects, for which reason adjustments of one object have no direct influence on the state of the other object. The next expansion stage is a digital shadow, where an automatic data flow from the physical object to its digital coun­terpart exists but not vice versa. The consequence is an automatic adaption of the virtual object's state if the physical object has undergone changes. The highest level of data integration, namely the digital twin, assumes a bidirectional automatic data exchange between both objects. It fol­lows a mutual automatic update of the respective counterpart (Uhlenkamp et al. 2019; Kritz- inger et al. 2018).

Abbildung in dieser Leseprobe nicht enthalten

Figure 2: Differentiation of digital model, digital shadow, and digital twin according to Uhlenkamp et al. (2019) and Kritzinger et al. (2018)

Characteristics

As introduced by Grieves, the digital twin can be characterized by the physical reality, the vir­tual reality and the (bidirectional) data link to exchange data between both realities: The main element of the physical reality is the physical system¹, i.e., the to-be modelled real-world arte­fact. The scope of this system in the respective applications depends on the determined bound­aries. Therefore, the subject of interest couldbe a robotic arm, a machinery, a whole factory or even a natural object such as a human being. The physical system is embedded into a physical environment, which mostly influences the defined scope of interest. The delineation varies from clear, e.g., regarding a single robot, to complex in form of a room in a building. The type of operation and expression of the physical system within the environment is exhibited through physical processes. In case of a robot, this can be gripping, cutting etc. (VanDerHorn and Ma­hadevan 2021; Jones et al. 2020).

The virtual reality can be decomposed into five elements. Depending onthe concrete use case, an “idealized representation of physical reality” (VanDerHorn and Mahadevan 2021, p. 3), i.e., a certain focus and degree of abstraction, must be selected. The system parameters and states are another element, which can be measured in the physical system or derived based on col­lected data. Furthermore, the virtual system and environment are the actual representations of its physical counterparts. Finally, virtual processes are the last component in the virtual reality. They “describe how the virtual system expresses itself atthe level of abstraction chosen for the virtual representation” (VanDerHorn and Mahadevan 2021, p. 4). Mostly, computational mod­els are applied to simulate how the system undergoes changes (VanDerHorn and Mahadevan 2021; Jones et al. 2020).

The data link consists of the physical-to-virtual, that means the transmission of collected data from sensors, manual input etc. to the virtual reality, and the virtual-to-physical data connection, which represents the feedback in form of general data, decisions, or actions back to the physical system. In a sophisticated digital twin application, the data fusion, i.e., the combination of var­ious data sets to deduce previously unavailable or more elaborated information, is also part of the data link (VanDerHorn and Mahadevan 2021; Jones et al. 2020).

The concept of data communication can also be referred to as digital thread. Concretely, a dig­ital thread is “a communication framework that allows a connected data flow and integrated view of the asset's data throughout its life cycle across traditionally siloed functional perspec­tives” (Leiva 2016). Thus, the connection of heterogeneous data sources and elements is en­sured - either within the digital twin or as standalone concept (Barricelli et al. 2019).

Finally, Figure 3 illustrates which processes run in the respective components of a general dig­ital twin: First, the physical system is operating on a regular basis and corresponding data are collected, e.g., via data measuring devices. Those data are interpreted and inter alia enriched by data fusion, which leads to an update of the virtual representation. Then, for instance through computational models such as simulations or predictions, the new state is analysed, and decision are derived based on the results. The result is transferred to the physical reality, where the phys­ical counterpart performs the related action. Those processes usually run iteratively and simul­taneously (VanDerHorn and Mahadevan 2021).

Abbildung in dieser Leseprobe nicht enthalten

Figure 3: High-level processes in digital twins by VanDerHorn and Mahadevan (2021, p. 3)

2.1.2 Underlying Technologies

The digital twin concept is based on and has several intersections with different modern tech­nologies. First, the digital twin application must collect, transmit and store data. The data col­lection is performed by sensors, cameras, scanners, radio-frequency identification (RFID) etc.

Then, data must be transmitted, usually via the internet in form of 4G, 5G or Wi-Fi, whereby especially the emerging 5G format can be seen as enabler for an almost real-time communica­tion. In this context, the internet of things (IoT) plays an auxiliary role through the internet­based interconnection of several objects. The data transmission is flanked by different proto­cols, e.g., the Hypertext Transfer Protocol (HTTP), to determine standards for the communica­tion. Further technologies are involved regarding the data fusion and data mapping, which is often implemented by means of XML (Liu et al. 2021b; Qi and Tao 2018).

The data storage in digital twins is particularly influenced by Big Data challenges such as the vast volume of sensor data. Another aspect is the cloud technology, which facilitates a cross­functional and ubiquitous accessible storage. Further information and a more detailed view on the data storage in digital twins is presented in 2.1.3.

The collected and stored data can be used for subsequent processing. On the one hand, the static and dynamic data are used to build and update the virtual model. Hereby, CAD or semantic web languages are possible options. Furthermore, human machine interfaces can be used to display additional data. On the other hand, virtual data can be generated and deduced based on the existing data sets. Thereby, virtual data include “model parameters and data of simulation, evaluation, optimization and prediction” (Tao and Zhang 2017, p. 20424). The application of the corresponding technologies is therefore self-evident (Tao et al. 2019; Liu et al. 2021b).

If the application entails the requirement to execute certain actions based on the decisions or findings of the virtual reality, then the physical counterpart needs actuators to perform those tasks. By considering new digital twin developments, edge computing and machine learning are mentioned in the literature as important future enhancements of respective applications. Hereby, edge computing facilitates data processing directly at the physical system, while the self-learning and self-adapting characteristics are fostered by machine learning (Qi and Tao 2018; Jones et al. 2020).

2.1.3 Data Storage in Digital Twins

Relevance and types of data stores

“There is no digital twin without data” (Wilmes 2019). Hence, the data storage is essential for a sophisticated digital twin as well, especially to perform auxiliary functions. This includes inter alia the fault analysis based on historical data, the exertion of prediction or decision models, or the (autonomous) optimization of the physical system, object, or process (Zibuschka et al. 2020; Kapteyn et al. 2020).

The actual data storage in digital twins can have different manifestations. The selection includes relational and non-relational databases, data lakes and data warehouses, each offline or in the cloud, as well as blockchain-based storage means. The decision, which type of data storage to use and how to configurate them, is to be taken based on the individual requirements of the use case.

The relational database is described as a “collection of data items with pre-defined relationships between them” (AWS 2021), which are organized in form of tables. In contrast, non-relational databases do not apply the tabular schema of most relational databases, but a data type specific design adapted to the related demands upon the data storage. Hence, non-relational databases are employed in order to store semi- or unstructured data (MongoDB 2021). Data lakes are capable of storing structured, semi-structured and unstructured data in its raw format. They can scale efficiently as well, especially since data will only be further structured in case of a subse­quent usage (Al-Ali et al. 2020). Furthermore, date warehouses represent databases for inte­grated data, that means they “can aggregate data over multiple sources like data lakes; however, the data must be organized and structured before storage” (Al-Ali et al. 2020, p. 6). The usage of blockchains for the data storage, especially in digital twins, is still in its infancy (Suhail et al. 2021b). Corresponding information are presented in-depth in chapter 2.2.1 and 2.3.1

Challenges

Independent from the type of data storage, certain challenges occur regarding the storage in digital twins, especially in the Big Data context (Suhail et al. 2021b). This addresses, contingent on the type of application, especially the following four dimension: volume, variety, velocity and veracity (Yaqoob et al. 2016). Thus, the massive data entail an increasingly difficult data collection, storage, management and analysis (Qi and Tao 2018).

First, the volume of collected and transmitted data in both, digital twin applications and other IT-related areas, is increasing exceptionally. The immense scale of data is related to several reasons. For instance, digital twin applications involve more and more sub objects with a grow­ing number of sensors. Furthermore, through high-quality images, videos etc., the amount of data experiences a strong surge as well (Yaqoob et al. 2016). As illustration, the yearly data volume of IoT connections worldwide is expected to grow between 2019 and 2025 by about 300% to 73.1 zettabytes (International Data Corporation 2020).

Second, the variety of data is very high in many digital twins as well. Several different data types, reaching from structured to semi- and unstructured data, ought to be collected and stored. This includes sensor values such as temperature, motion or pressure data, text documents, re­ports, images, videos or other data files (Dietz et al. 2019).

Third, another challenge in digital twin applications is the velocity of the data threads. In order to ensure a real time representation of the physical counterpart, a corresponding near real time data collection by sensors and transmission must be established. Data handling means and the data storage must be adapted to fulfil the high processing speed because thereby also the quan­tity of data is increased (Dietz et al. 2019).

Fourth, the term veracity addresses challenges regarding the uncertainty and trustworthiness of data as a consequence of inconsistencies, incompleteness, missing quality etc. By considering digital twins, this represents a major challenge since many (untrustworthy) parties might be involved in the data generation plus data sources could even be completely unknown (Lu­koianova and Rubin 2014). Related topics in this field are the data transparency respectively privacy and the traceability of data in digital twins (Suhail et al. 2021a).

However, the minimally required data volume or velocity is strongly dependent on the applica­tion and often entails a trade-off about the concrete extent. This is related to the drawbacks of both, a low and high data volume respectively velocity. For example, one the one side by relying on too little data, inaccurate results and predictions may follow. On the other side, too many data may “pose decision paralysis due to information overload” (Suhail et al. 2021a, p. 7).

2.1.4 Benefits and Application Areas

Benefits

The digital twin provides several benefits, which can be outlined and classified along its product life cycle. They appear to a different degree depending on the use case, however, are mostly valid on a generic and cross-application fundament.

During the design phase of the corresponding physical object, a digital twin enables the creation and evaluation of different design alternatives on a virtual basis. Therefore, product and regu­latory requirements can be addressed, and engineers can perform tests to simulate unexpected situations in order to eradicate product deficiencies. Through the improved risk assessment and more efficient design processes, an enhanced time to market and error-reduced commissioning and operation can be achieved (Luber and Litzel 2018; Groesser 2018).

Similar to the previous phase, digital twins show similar advantages in the manufacturing phase. Here, the digital twin can help to attain a higher efficiency, quality, and yield through the analysis of collected data (Groesser 2018).

Many benefits of a digital twin are revealed in the operation phase. For instance, a high trans­parency and better coordination among different parties result from the real time monitoring of the corresponding application. By analysing the object conditions, workflows can be improved and optimized in terms of its efficiency, effectiveness, costs etc. Through the simulation of potential operations, the digital twin can deduce recommendations for actions without inter­rupting the physical procedures. Furthermore, lower maintenance expenses are reached by means of predictive maintenance and the related remote control (Groesser 2018; Boschert and Rosen 2016; Jones et al. 2020).

Finally, in the recycling phase, the digital twin fosters a structured and frictionless replacement planning and facilitates the upcycling, i.e., the reusage of certain parts because precise infor­mation about exploitable components might be available (Groesser 2018).

Application areas

The most popular field in which digital twins are applied is manufacturing. Hereby, the moni­toring of production machines, production processes or actual products represents a common application. Through the obtained information, procedures are optimized and e.g., supply chain movements can be simulated, predicted, and visualized. Furthermore, the tracking of products allows the firms not only to save money and time, but they can also ensure a higher level of proven sustainability through the precise traceability (Barricelli et al. 2019; Fuller et al. 2020; Sallaba et al. 2018).

The smart city context holds many potential application areas as well, especially regarding in­frastructure and services. This is related to the increasing number of smart cities and connec­tions among communities, objects, and procedures. Examples are the continuous surveillance and subsequent optimization of buildings, energy sources such as windmills, or general infra­structures. Similar to the manufacturing area, predictive maintenance can be applied to those utilities. Furthermore, service-related examples in smart cities are the management of transpor­tation and traffic, consumer-based services such as virtual assistants, or the cultivation of agri­cultural land (Barricelli et al. 2019; Fuller et al. 2020; Sallaba et al. 2018).

Healthcare, which is assignable to smart cities, is another major application area of digital twins. This includes for example the planning and control of surgeries, the simulation of the impact of drugs, or the forecasting demand planning of treatments. In research projects, there are even experiments to reflect the human being in digital twins (Barricelli et al. 2019; Fuller et al. 2020; Sallaba et al. 2018).

2.2 Blockchain

2.2.1 Foundations of the Blockchain Technology

Origin and definition

The first known examination and publication of blockchain-related technologies about digital documents was made by Haber and Stornetta in 1991. In their research paper named How to Time-Stamp a Digital Document, they discussed the usage of computer servers to “timestamp and link digital documents as a chain with pointers attached to the data in each document” (Liu et al. 2020a, p. 395). The pointer invalidity in case of data changes guaranteed forgery-proof documents after the server had signed them. Along with their basic research about hashing or distributed trust, those advances represent fundamental pillars of modern blockchains (Liu et al. 2020a; Manu et al. 2021; Haber and Stornetta 1991).

However, a broad awareness and relevance of this topic was obtained in 2008 through a paper about the cryptocurrency Bitcoin by a group pseudonymously named Satoshi Nakamoto. The underlying technology of Bitcoin is blockchain, for which reason Nakamoto is considered as the accredited originator behind the blockchain and distributed ledger technology (Manu et al. 2021). Thereby, blockchain is one type of a distributed ledger technology, which relies on the technical specification of storing transaction data in a sequence of blocks. Correspondingly, distributed ledgers in general do not require chains, but can be realized by means of different protocols such as a directed acyclic graph or Hashgraph (Dolenc et al. 2020; Witt 2021). As illustrated in Figure 4, distributed ledgers are one manifestation of distributed databases, that are “characterised by its consensus-based validation process” (Ugarte and Luis 2018, p. 2).

Abbildung in dieser Leseprobe nicht enthalten

Figure 4: Classification of blockchains inspired by Hileman and Rauchs (2017, p. 23)

To provide a common understanding and distinguishment of the blockchain technology, in this work the blockchain is defined as follows:

“A blockchain is a distributed database, which is shared among and agreed upon a peer-to-peer network. It consists of a linked sequence of blocks, holding timestamped transactions that are secured by public-key cryptography and verified by the network community. Once an element is appended to the blockchain, it can not be altered, turn­ing a blockchain into an immutable record of past activity” (Seebacher and Schüritz 2017, p. 14)

Apart from the widespread usage of the blockchain technology in the financial sector, e.g., regarding cryptocurrency, the technology is more and more adapted for use cases concerning supply chain transparency, identity management, fraud prevention in online voting or money laundering, or in the area of internet of things (Manu et al. 2021).

Operating principle

As simplified depicted in Figure 5, a blockchain consists of data sets that are orchestrated in form of a chain of blocks which function as data packages. A block is composed of a timestamp, several transactions, the hash value of its preceding block, and a nonce, i.e., a byte number to validate the hash. The first block is named genesis block and has no predecessor. Hashing is a core element in blockchains and enables thecoding of transactions through the transformation of random strings into uniform, uniquely mapped codings. Another major element of the tech­nology, which is described below, is the consensus mechanism to validate the correctness of transactions (Prinz et al. 2018; Nofer et al. 2017; Manu et al. 2021; Zheng et al. 2018).

Abbildung in dieser Leseprobe nicht enthalten

Figure 5: Example of a blockchain structure (Zheng et al. 2018, p. 355)

As illustrated in Figure 6, at first a transaction, e.g., the registration of collected IoT data or documents, is generated from a sender and digitally signed. Then, the transaction is forwarded to the network and distributed to the involved nodes. The new transaction is not automatically stored in the distributed ledger but requires a successful examination through a consensus build­ing within the network. The underlying consensus mechanism can take several forms (cf. 2.2.2) and is “the process in which a majority (or in some cases all) of network validators come to agreement on the state of a ledger. It is a set of rules and procedures that allows maintaining coherent set of facts between multiple participating nodes” (Swanson 2015, p. 4). Thereafter, the consensus-based transactions are stored in a block and transformed into a uniform format via hash functions. Through the linkage with the existing blocks, the chain is extended and hence “represents a complete ledger of the transaction history” (Nofer et al. 2017, p. 183).

Furthermore, fraud can be prevented as a consequence of the unique hash values, as changes would have a direct influence on the hash value. Because also the addition of new blocks re­quires a consensus building, an irreversible and validated chain of blocks results. In order to maintain data persistence, the chains of blocks are replicated within the network (Nofer et al. 2017; Prinz et al. 2018; Swanson 2015).

Abbildung in dieser Leseprobe nicht enthalten

Figure 6: Working principle of a blockchain (Prinz et al. 2018, p. 313)

Key characteristics of blockchains

The blockchain technology is related to some key characteristics in comparison to other means of data storage. Depending on the actual form and manifestation of the blockchain, these char­acteristics emerge with a varying degree of intensity, however, it is commonly accepted that they are strongly linked to the technology:

One of the most critical characteristics of blockchains is d ecentralization. The lack of central­ized control follows from the missing administrative authority to inspect the structure or storage process. The blockchain utilizes resources of any node of the network and data can therefore be accessed and copied by everyone directly via the internet. Thus, the potential issue concerning a single point of failure as well as certain latency problems can be mitigated through the elimi­nation of many-to-one communication mechanisms (Atlam et al. 2018; Ramasamy and Kadry 2021; Sultan et al. 2018).

The blockchain represents a record of transactions, which is immutable and tamper resistant. Once a block is inserted to the blockchain based on a consensus, it cannot be altered, updated, or deleted and is therefore persistent. Besides, no transaction can be added to the distributed ledger without the consent of (the majority of) nodes in the network (Zheng et al. 2017; Rama- samy and Kadry 2021; Atlam et al. 2018).

Another characteristic of the blockchain technology is anonymity. Through the possible gener­ation of one or various addresses by each user, they can avoid the disclosure of their real iden­tity. Therefore, no central authority stores personal information about any user. However, the blockchain “cannot guarantee the perfect privacy preservation due to the intrinsic constraint” (Zheng et al. 2017, p. 558), i.e., the balances for the addresses of the transaction and the respec­tive transaction values are visible within the network (Zheng et al. 2017; Zheng et al. 2018).

Although the blockchain ensures a certain level of anonymity, the changes in a blockchain are visible and can always be traced back. The result is a stable blockchain which provides trans­parency regarding their underlying transactions. They can easily be audited and verified be­cause transactions are stored with a timestamp and users can comprehend previous records by observing different nodes in the blockchain network (Zheng et al. 2018; Sultan et al. 2018).

The removal of a central authority leads to an improved safety as a change of the properties is prevented. Another aspect in terms of safety is the usage of cryptography. The decentralized nature of the blockchain and the irreversible hashing of every data packet represent an addi­tional security layer - providing “an extremely mixed mathematical method that works as a firewall for attacks” (Ramasamy and Kadry 2021, p. 46). Furthermore, a single point of failure is inhibited. Moreover, the change or damage of any data in the blockchain would require the modification of every hash ID of every block, which leads rapidly to millions or billions of nodes that needs to be manipulated by the attacker. This makes the unauthorised altering of data infeasible (Ramasamy and Kadry 2021; Atlam et al. 2018; Seebacher and Schüritz 2017)

Consensus

The last key characteristic covered in this work is consensus. As described above, the trust­based peer validation is essential in blockchains to make network and storage decisions. Despite the lack of knowledge and trust towards other nodes, they can rely on the consensus mechanism, whose decision can be attained quickly (Sultan et al. 2018; Ramasamy and Kadry 2021).

2.2.2 Different Blockchain Configurations

Type of blockchain

The blockchain technology can be roughly subdivided into three different types, namely pri­vate, public, and consortium blockchains. This differentiation addresses the different levels of access restrictions, especially for reading data. Public and private blockchains, which are con­sidered as the most fundamental types of blockchain, are placed at the opposite ends of the corresponding continuum (Niranjanamurthy et al. 2019; Rio 2020).

Public blockchains allow the read access by any node and user on the internet without re­strictions. The lack of a central authority leads to a decentralized approach. The result is a high level of information transparency and data immutability, however, paired with sacrifices in terms of efficiency and additional cryptographic endeavours. The majority of blockchains for cryptocurrencies are public, e.g., Bitcoin or Ethereum. In contrast, private blockchains are con­trolled by a single organization in a decentral manner, i.e., not every node can participate in the blockchain. Even though read permissions can be public or restricted (determined by the au­thority), transactions can only be validated by the organization. The consequence is a highly efficient and flexible type of blockchain, but associated with the risk of hampering its immuta­ble character (Zheng et al. 2017; Xu et al. 2017; Wegrzyn and Wang 2021).

A consortium blockchain is considered as a partially decentralized blockchain, which is con­trolled by a selected set of nodes, mostly distributed across multiple businesses or organizations. The access to reading mechanisms can be provided to a restricted set of nodes or everyone. An example for a consortium blockchain is Hyperledger (Zheng et al. 2017; Niranjanamurthy et al. 2019).

Permissionless vs. permissioned

The next differentiation regarding the configuration of a blockchain aims at different rights to validate new transactions, i.e., to build and add new blocks to the blockchain respectively write into the blockchain. As the name suggests, permissionless blockchains allow every participating client to publish blocks without the need of a permission from any authority. As anyone can participate, attempts of malicious interactions ought to be prevented by means of agreements or a consensus across multiple parties, e.g., in form of a Proof-of-work (PoW) or Proof-of-Stake (PoS) mechanism.

A permissioned blockchain has only a defined list of nodes with related identities that are al­lowed to write to the blockchain, i.e., they must be authorized by a certain type of authority. Hence, the reading access and issuing of transactions can be regulated as well. Consensus mod­els can also be applied, however, in contrast to a permissionless blockchain, in a more cost­effective way. In general, permissioned blockchains are related to a higher level of trust in the validator, a better scalability and only a partly realized decentralization (Yaga et al. 2018; Kudra et al. 2017).

Ledger structure

Basically, the ledger structure can be subdivided into four different forms according to Xu et al. (2019). For example, despite certain forks exist that are important for the operation of the blockchain, Bitcoin respectively Ethereum are each considered as a global list of blocks. This structure promotes the key characteristics of a blockchain ideally. Alternative ledger structures are a global directed acyclic graph (DAG) of blocks respectively of transactions (e.g., IOTA). All of the abovementioned ledgers have one global transaction history. In contrast, restricted shared ledgers (e.g., Corda) represent many small ledgers that are only distributed among au­thorized entities. Although the overall structure is aligned to a global graph of transactions, most entities perceive it as a “collection of small ledgers” (Xu et al. 2019, p. 51). In contrast to the classical global chain, the DAG and restricted shared ledgers have a higher flexibility, per­formance and cost efficiency but do address the key characteristics less adequately. Although alternative ledger structures such as the DAG are not considered as blockchains in the classical sense, they are still respected in the varieties of blockchains (Xu et al. 2019).

Consensus mechanism

The selection of the consensus mechanism has an influence on the scalability, security, latency, and throughput of the blockchain operation. A well-known protocol, which is used in the Bitcoin network, is PoW. Hereby, various nodes calculate a solution to a mathematical problem for each new block based on its predecessor. The node which solves the puzzle first, can add the new block and receive a reward in return. This makes PoW a computationally intensive and low-performant protocol. (Xu et al. 2019; Irannezhad 2020).

Another famous consensus protocol is PoS. Here, participants who hold tokens (e.g., specific currency) temporarily put their tokens in a locked smart contract. Simultaneously, they validate blockchain transactions and earn incentives in return based on their relative share of tokens. This mechanism is more cost-efficient and flexible compared to the PoW. Similar to PoS is the protocol delegated Proof-of-Stake. In contrast to PoS, there is a representative democratic con­sensus, i.e., delegates are chosen to build and confirm blocks, which leads to fewer nodes and a faster validation of transactions (Zheng et al. 2017; Irannezhad 2020).

The Practical Byzantine Fault Tolerance (PBFT) protocol is rather applied in permissioned blockchains since all participants have to agree upon the concrete line-up of participants. Sim­ilar to PoW, validators verify the requested block in a low-trust context. Every node issues a public key. As transactions arrive, they are signed by several nodes in order to confirm them. This process ends with a consensus in form of a minimum proportion of identical responses from the nodes. Despite a proficient level of throughput, latency, and cost-efficiency, the PBFT shows a bad scalability and low level of anonymity as a consequence of a relatively small num­ber of participating entities (Xu et al. 2019; Irannezhad 2020).

Another consensus mechanism is the Proof of Elapsed Time (PoET), which uses a lottery mech­anism combined with the waiting time of trusted functions to determine the node to publish blocks. Thus, through a trusted execution environment, computation costs are decreased. Be­sides the above mentioned mechanisms, there exist a large variety of additional consensus mechanisms, e.g., Proof-of-Authority (PoA) or Bitcoin-NG, which are covered in many scien­tific publications in-depth (Lang and Karlstetter 2017).

Block configuration

The configuration of the blocks in a blockchain can have a direct influence on the transaction processing speed (and therefore on the scalability) by adjusting the block size and the genera­tion frequency. The block size is represented for instance by the actual data size or the com­plexity of the inherent transactions depending on the blockchain. Larger blocks result in a higher throughput of transactions, however, are related to a lower speed of replication and trans­action processing - leading to a trade-off between those characteristics. On the other hand, by simplifying the mining difficulty, the latency and throughput can be enhanced but the emer­gence of additional forks and a higher number of confirmation blocks² is fostered (Xu et al. 2019).

On-chain vs. off-chain

Due to the limitation of storage and computational effort in a blockchain and the related oper­ating costs, a significant configuration option is the selection, which data and computations are moved off-chain. The actual storage of data can either take place on-chain - embedded into transactions or in combination with smart contract - or off-chain in a cloud or network. This selection has a direct influence on several blockchain-related properties. Off-chain storage en­ables a higher performance and flexibility but concurrently undermines the key characteristics such as immutability or transparency. Furthermore, the computations in a blockchain network can be performed off-chain via a third party respectively private cloud, or on-chain via smart contracts etc. The chosen design has similar influences on the blockchain characteristics as the form of storage (Xu et al. 2017).

Additional configuration options

Not only the blockchain inherent configuration must be selected, but also the actual blockchain platform. Apart from the most famous blockchain platform Bitcoin, Ethereum experiences a growing interest due to its additional non-financial application areas. Further platforms are for instance Hyperledger or Guardtime. The selection of the blockchain platform entails automati­cally certain configuration options (Burgwinkel 2017).

If the usage of an existing blockchain is refused, the combination with an existing system to an “auxiliary blockchain” (Xu et al. 2019, p. 54) is an alternative. The form of the new blockchain can either promote security or scalability-related aspects. For scalability, “multiple blockchains can be used to isolate information of separate concerns” (Xu et al. 2019, p. 55), e.g., via several private blockchains, mini-blockchains or sidechains³. In contrast, security is fostered by using a new blockchain that is “aligned with public blockchains, utilizing existing infrastructure, re­sources, and trust” (Xu et al. 2019, p. 54). One related example is the merged mining (Xu et al. 2019).

2.3 Blockchain-based Digital Twins

2.3.1 Rationale for Combining Blockchains and Digital Twins

Essential for the success of the digital twin concept are data. Those data are interchanged be­tween the virtual and real world, whereby often several different parties are involved. The se­curity, transparency, and trust in platforms and applications for digital twins, especially in case of unknown partners, is a fundamental prerequisite for the usage of this technology. The block­chain technology represents one way for providing an environment to foster a trustworthy data communication and storage (Raj 2021; Miskinis 2018).

Although the interplay of the blockchain technology and digital twins and the corresponding research is still in its infancy, Deloitte and RIDDLE&CODE describe the blockchain technol­ogy as “potentially the most suitable and efficient way to generate, monitor and exchange digital twins” (Sallaba et al. 2018, p. 4). Especially use cases with different untrusted participants and critical data seem suitable. Examples are the application of the two technologies for the product life cycle and supply chain tracking e.g., to ensure the compliance of sustainability criteria for each party or prevent counterfeits (Raj 2021; Suhail et al. 2021b). The increasing interest in the combination of the blockchain technology and the digital twin is further reflected in Figure 7, as there is a growing amount of research papers in the last few years.

Abbildung in dieser Leseprobe nicht enthalten

Figure 7: Research trends of blockchain-based digital twins in industry according to Suhail et al. (2021b, p. 8)

2.3.2 Benefits

The key functionalities of blockchains (described in the chapter Foundations of the Blockchain Technology) lead to a row of benefits, that are beneficial for the usage of digital twins. First, the identity of a digital twin can be confirmed and legitimated through immutable digital cer­tificates, which hampers fraudulent accesses. Furthermore, there is a trusted data communica­tion and coordination in digital twins since several nodes in the network can control the data until a consensus for the final storage is reached. Due to the cryptographic means and control mechanisms, a high level of accountability and security of the digital twin data can be attained. Therefore, attacks from hackers can be significantly mitigated. Another benefit is the transpar­ency, which results from the possibility to trace back past transactions (Yaqoob et al. 2020; Sallaba et al. 2018; Suhail et al. 2021b).

The performance of decentralized and immutable data transactions can also be seen as major benefit of the blockchain through its distributed ledger and the underlying hashing mechanisms for data storage. Further advantages are the diverse access possibilities for different parties due to the distributed character, i.e., the data storage is accessible from multiple sites but is simul­taneously controlled automatically and safely. Finally, the complementary application of smart contracts enables the automation of different procedures and scenarios which could enrich the range of appropriate use cases of digital twins (Nielsen et al. 2020; Yaqoob et al. 2020).

2.3.3 State of the Art

The fact, that blockchain-based digital twins are still in its infancy, is indirectly confirmed by the relatively small research basis. There exist especially theoretical discussions regarding the interplay of the blockchain technology and digital twins. Yaqoob et al. (2020), whose insights contribute significantly to chapter 2.3.2 and who are considered as one of the most important papers in this field, outline benefits and several open barriers for the combination of blockchains and digital twins. Another theoretical work addresses advantages and use cases of blockchain­based digital twins as well as the related key features of the blockchain technology (Raj 2021). Besides the work by Götz et al. (2020), which investigate the interoperability and integrability of the blockchain technology in digital twins for the asset life cycle management, further pub­lications place its focus stronger on one technology (either blockchain or digital twin) and dis­cuss certain interfaces (e.g., Leng et al. 2020a).

Suhail et al. (2021b) published the first paper that explores current literature regarding block­chain-based digital twins approaches comprehensively. In their work, they identified nine pa­pers (depicted beneath), in which the authors realized corresponding approaches to a different level. Furthermore, they identified related issues and future challenges.

One of the abovementioned approaches was developed by Angrish et al. (2018), which aims at the decentralized handling of manufacturing information from various participating entities via the blockchain technology. Therefore, they used the Ethereum blockchain as well as the related smart contracts to enable an automated, event-triggered interchange of commands between ma­chines. Furthermore, Angrish et al. (2018) examined PoW and PoA as consensus mechanism to weigh them up depending on the underlying application. Despite the lack of an implementa­tion, they proposed storing data off-chain to address a certain level of data volume and velocity.

Dietz et al. (2019) developed a concept for the digital twin data sharing based on the distributed ledger technology with an emphasises on security-related aspects such as traceability or access control. Although they did not specify the type of blockchain or consensus mechanism, an off- chain storage, e.g., via distributed hash tables⁴, is strongly suggested. In this context, they high­light the variety and velocity of digital twin data that must be stored. Smart Contracts are used in their work as well, primarily in order to record authorization data from all participating enti­ties.

Similarly, Hasan et al. (2020) propose in their work a blockchain-based creation process of digital twins to enable the immutability, traceability etc. of transactions. Hereby, they also sug­gest using Ethereum, Smart Contracts and an off-chain storage based on the InterPlanetary File System (IPFS). However, the prototype is discussed primarily on a theoretical basis and lacks details about the implementation.

In contrast, Leng et al. (2020b) and Zhang et al. (2020) used in their approach Hyperledger as blockchain. Nonetheless, the findings indicate that the majority of researchers use sequential chain-based blockchains with a high level of maturity. In their prototype regarding manufac­turing system digital twins, both author groups use the consensus mechanism Cross Fault Tol­erance with the aim of guaranteeing a high data throughput (Suhail et al. 2021b).

Next, Putz et al. (2021) developed a prototype named EtherTwin, which is, at the beginning of 2021, regarded as the only non-rudimentary prototypical implementation in this field⁵ (Suhail et al. 2021b). EtherTwin is a “blockchain-based Decentralized Application (DApp) for secure information management of Industry 4.0 assets using Digital Twins” (Putz et al. 2021, stated in the highlights on ScienceDirect). As the name suggests, they used Ethereum and implemented an off-chain storage, encryption, and access control thoroughly.

Finally, three further blockchain-based digital twin approaches are present in the academic lit­erature (Huang et al. 2020, Khan et al. 2022, Mandolla et al. 2019). They investigate the creation of a digital twin of different products through the application of the blockchain technology. However, the papers did not cover aspects related to data storage and placed their focus rather on the theoretical discussion than on realization details. Only Khan et al. (2022) specified the type of underlying blockchain, namely a new variant named twinchain, that fulfils requirements for velocity and quantum-resilience (Suhail et al. 2021b).

All in all, the current literature concerning blockchain-based digital twins demonstrates the early research stage. The number and extent of (partial) prototypical implementations are still low and significant properties such as the granularity of data and the data fusion are not covered in the respective approaches (Suhail et al. 2021b). However, the increasing interest and engage­ment of the industry in related research projects, e.g., by Bosch Rexroth and Phoenix Contact, indicate good progress in the near future (Heißmeyer and Kalhoff 2020).

3 Methodology

Section 3 provides an overview of the underlying methodology in this work. The mixed-method approach is comprised of four components. First, the systematic literature review and taxonomy development method for the classification of data-related properties in digital twins are speci­fied in detail. Second, the procedure for the elaboration of requirements for the data storage in digital twins via expert interviews is explained. The third part addresses the procedure for the evaluation of the suitability of the blockchain for the data storage in digital twins. Finally, the approach for the development of the design decisions for the blockchain-based storage of digital twin data is delineated.

3.1 Systematic Literature Review and Classification

Systematic literature review

The first major stage of this research forms a systematic literature review, which represents the basis for the development of a scheme for the classification of data-related properties in digital twins. Therefore, this work explores several realizations or applicatory frameworks of digital twins in science in order to extract information about the data storage, data-related mechanisms and data itself. The approach for the systematic literature review is derived from the methodo­logical guidelines of Vom Brocke et al. (2009). Vom Brocke et al. (2009) propose a five-step approach including the specification of the review scope (step 1), the conception of the research area (step 2), the search process (step 3), the subsequent analysis (step 4) and the definition of a research agenda (step 5). Hence, at first the review scope was defined as scientific papers about applications of digital twins covering data-related topics as integral element. Second, a working definition - equalling the final definition presented in section 2.1.1 - was provided to conceptualize the domain. As described below, the third and fourth step, namely the literature search and analysis constitute the largest part of the literature review.

As depicted in Figure 8, the search process was initiated by the selection of a database respec­tively search engine. Google Scholar was chosen as search engine mainly for two reasons, namely its broad coverage - reflected in the indexation of most major publishers and databases such as IEEE or Science Direct - and the simultaneous currency (Jacso 2008; Vine 2006). Sub­sequently, an adequate search string was determined and applied solely to the title of the docu­ments in order to match the review scope as precisely as possible. The first component of the search string is “digital twin”. The inclusion of further synonyms, e.g., virtual representation has been disregarded because of the common useof the term digital twin. In contrast, the second part of the search string incorporates different substitutes of “use case” to address a broader bandwidth of realizations of digital twins and to extend the number of search results to 676 papers. With the aim of reaching a for this work adequate scope, the research results were au­tomatically sorted by relevance and subsequently limited to the first 180 papers. In addition, the fact, that the sorting algorithm considers the citation count as highest weighted factor, led to a direct focus on the most popular and renowned publications in this field (Beel and Gipp 2009).

Abbildung in dieser Leseprobe nicht enthalten

Figure 8: Literature search process including filtering

The next decimation to 81 instances was attained through the first content-oriented filtration, at this point based onthe respective titles and abstracts. Especially documents with a narrow focus on mathematical machine learning models or a lack of emphasis on general data-related issues and the digital twin itself were excluded. Another reason for discarding certain papers was their concentration on theoretical discussions and trade-offs instead of determining actual manifes­tations in terms of data storage, collected data, data processing steps or applied technologies.

The second content-based filter targets the document access, quality, and potential duplicates. By means of a broad scan on different databases and requests to every author, it was able to get access to several papers with a previously restricted access. However, nine documents remained without access or with relatively high paywalls. Besides, seven documents were excluded since they were not available in English, had duplicates or exhibited an insufficient quality in the form of published Master Theses, which led to 65 remaining papers.

The final filtration consisted of a full text review, in which the papers were re-examined for a satisfactory level of information content regarding the abovementioned data-related properties. In conclusion, the result consists of 42 papers, which were analysed with the aim of classifying the digital twin-inherent data properties. Figure 9 demonstrates the corresponding assurance of a certain level of quality and academic rigor as only conference and especially journal papers were considered. Furthermore, the temporal distribution with a main emphasis on papers pub­lished after 2018, illustrates the topicality of the selected references and simultaneously the novelty of the related research field.

The fifth and final step of the systematic literature review approach by Vom Brocke et al. (2009) was realized by describing related future research topics in chapter 6.3.

Abbildung in dieser Leseprobe nicht enthalten

Figure 9: Paper distribution by year of publication

Taxonomy development

For the purpose of classifying certain objects, a taxonomy was selected as well-suited approach since it facilitates a better understanding and further analysis of sophisticated domains (Nick­erson et al. 2009). Furthermore, the reduction of complexity as well as the identification of differences and similarities across several digital twin applications represent substantial bene­fits of the usage of the taxonomy (Bailey 1994). Finally, a taxonomy is mostly divided into several layers, each consisting of a dimension and “their modes of occurrence, called charac­teristics” (Nagel and Kranz 2020, p. 207). The chosen procedure in this study was aligned with the taxonomy development method by Nickerson et al. (2013) as depicted in Figure 10, which is considered as one of the most popular methods in this field.

[...]

---

¹ When referring to the physical system, object etc. in a general manner, other physical types of counterparts are included as well (e.g., process).

² The related modus operandi in most mainstream blockchains (i.e., those who use “Nakamoto consensus”, e.g., Ethereum or Bitcoin) to confirm transactions is as follows: The transaction “needs to be included in a block, which should be endorsed by dependent blocks, known as confirmation blocks” (Xu et al. 2019, p. 197).

³ “Sidechaining is a mechanism that allows tokens of one blockchain to be securely transferred and used in another blockchain; eventually, they can be moved back to the original chain securely” (Xu et al. 2019, p. 56).

⁴ “A Distributed Hash Table (short DHT) is a data structure to locate and distribute information (such as files) in peer-to-peer networks.” (Matthes 2020, p. 7).

⁵ The current status cannot be validated, as the remark was made in March 2021 in Suhail et al. (2021b).

• Hochschule: Universität Hohenheim
• Autor: Tobias Heinrich (Autor:in)
• Erscheinungsjahr: 2022
• Seiten: 128
• Katalognummer: V1314320
• ISBN (PDF): 9783346792686
• ISBN (Buch): 9783346792693
• Sprache: Englisch
• Schlagworte: blockchain-based, digital, twins, requirements, design, decisions, data, storage, Digital Twins