Digital Twin

The German government's Industrie 4.0 Platform defines the digital twin briefly as a „virtual digital representation of physical assets“. ^[30]

Fraunhofer IOSB defines the digital twin in more detail as follows:

„The Digital Twin is a concept that models products as well as machines and their components using digital tools, including all geometry, kinematics and logic data. A digital twin is the image of the physical ‘asset‘ in the real factory and allows its simulation, control and improvement.“ ^{[11] [9]}

Four essential components of a Digital Twin are represented by the following elements:^[7]​

• a digital/virtual representation​
• a (directional) information exchange between the physical and virtual world​
• a life cycle view​
• a composition of different models, some of which are suitable for simulation. ​

Taking into account the aforementioned components, it becomes clear that the digital twin is a virtual image of an object. The object can be tangible or intangible (e.g. product, system, process or service).^[6] Furthermore, it is emphasised that there is an exchange of data between the physical and virtual object and it is pointed out that the digital twin accompanies the physical object throughout its entire life cycle and digitally reflects it across different stations (development, production, use, etc.) ^[8]. The last component illustrates that the digital twin consists of different models and, if necessary, is created with different digital tools on suitable platforms.^{[6] [9]}

In summary, it is possible to state the following about the digital twin:

The digital twin is a digital representation of tangible or intangible objects (e.g. devices, machines, systems, services) and is created with several digital tools. The digital representation contains characteristics, states and the behaviour of an object or system and is fed by data that is exchanged between the physical and virtual world. Furthermore, the digital representation accompanies the physical object throughout its life cycle and during individual life cycle phases various models, information and data are linked with each other in the digital representation. This allows the physical object and its behaviour to be monitored, controlled, improved and simulated.^{[6] [7] [8] [9] [10]}

A digital model is a digital representation of an existing or planned physical object that does not use any form of automated data exchange between the physical object and the digital object (see Figure 1). The digital representation may contain a more or less comprehensive description of the physical object. At the same time, these models (simulation models, mathematical models, etc.) do not include any automated data integration. This means that all data exchange is done manually. A change of state of the physical object has no direct effect on the digital object and vice versa.^[21]

Figure 1: Data transfer in digital model

If there is an automated one-way data flow between the state of a physical object and a digital object, we can speak of a digital shadow (see Figure 2). This means that a change of state of the physical object leads to a change of state of the digital object, but not vice versa.^[21]

Figure 2: Data transfer in digital shadow

If data flows are fully integrated between a physical object and a digital object in both directions, we can speak of a digital twin (see Figure 3). In such a combination, the digital object could also function as a control instance of the physical object. There can also be other objects, physical or digital, that cause changes in the state of the digital object. A change of state in the physical object leads directly to a change of state in the digital object and vice versa.^[21]

Figure 3: Data transfer in digital twin

The digital twin creates the necessary transparency to virtually design, test and optimise products, processes and systems. New applications and business models, but also partnerships, can be tested digitally without having to build them in reality.^[20] This implies that digital twins have the ability to represent different information in a uniform format. However, digital twins are more than just data. They contain algorithms that accurately describe their real-world counterpart.^[25] At a high level of sophistication, the digital twin even gives feedback directly to its physical brother, and together they form a self-controlling and self-improving AI system. Information management is the key to providing relevant information to decision-makers in a timely and secure manner.^[20] The digital twin offers further advantages through simulation in the engineering phase. This improves collaboration between experts, as complex scenarios can be better explained through simulations. Furthermore, the application increases efficiency as well as the possibility to carry out a virtual commissioning. This shortens commissioning times. In addition, real-time monitoring makes it possible to respond quickly to changes in condition, which supports maintenance.^[7]

Positive synergies also result from digital twins on supply chains. By monitoring and simulating events (e.g. natural disasters, trade disputes, etc.) that affect supply chains, digital twins enable an early and proactive response. In addition, end customers also benefit from digital twins. For example, products can be visualised in 3D in advance for customers, creating a closer and more personalised customer experience.^[26]

In order to exploit the potential of a digital twin, it is essential to know its architecture. Due to the continuous and dynamic expansion of the scope of digital twins, it is necessary to extend the three-dimensional model proposed by Grieves (2002) into a five-dimensional model. This new model consists of the following five dimensions: physical objects, digital objects, services, data and connections (see Figure 4).^[29]

Figure 4: The five-dimensional digital twin model

1. The physical object is the basis of a digital twin and can be represented by a device or product. With the help of a sensor, information is gathered in form of data from the real world (e.g. measurement of physical and chemical quantities). The generated data is stored (some examples of storage methods are database and cloud storage) and transmitted between the physical and virtual world, creating a connection.^[29]

2. A digital object is developed from a physical object to reproduce the physical properties and behaviour of the object in the real world. A variety of models and simulation tools are required to create the virtual model.^[29]

3. Another dimension of the model proposed by Tao^[29] are the services. From the dynamic flow of data that is generated by this model, it is possible to perform several services such as: simulation, verification, optimisation, diagnosis and prediction .^[29]

4. Data is a very important factor in the architecture of the digital twin. A very transcendental issue is the origin of this data. The data can be obtained from physical objects, generated by digital objects as result of a simulation, provided by the services, supplied by experts in the subject or extracted from existing data.^[29]

5. The connections are the fifth dimension of this model. Digital objects are dynamically connected to their real-world counterparts to enable advanced simulation, operation and analysis. The connections between physical entities, virtual models, services and data, enable the exchange of information. As shown in Figure 4 there are 6 connections in this model.^[29]

The digital twin presents a wide range of application areas due to its characteristics and its great technical potential. The following table (see Table 1) shows the possible applications and the added value of this technology ^[5]

Table 1: Areas of application of digital twins ^[5]

So far, there is no standardised architecture for the digital twin. This implies that institutions and companies can develop individual digital twins, depending on the specific application and product. Another challenge is that the digital twin often only describes one lifecycle stage and does not continue through several lifecycle phases. Therefore, it is necessary for the digital twin to continue throughout the lifecycle of the physical counterpart. In order to achieve this, different companies must work together and cooperate. Otherwise, the potential of the digital twin will not be fully exploited.^{[7] [27]}

There are several challenges for the management of digital twins. These include the identification and data management of the product along the product life cycle, the creation of simulation models in different IT systems and the control of the huge amounts of data.^[9] The simulation models in particular form the basis for the diversity of a digital twin. This heterogeneity is one of the major challenges for the concept of the digital twin. In order to overcome this problem, a continuous flow of information between the heterogeneous models is necessary. Co-simulation and cognition are technologies that address this challenge of heterogeneity and are already established in many application areas. For co-simulation, different simulation tools have to be coupled, whereas cognition is about the ability to analyse and visualise information. However, these technologies require adaptations and extensions to meet the expectations of the digital twin.^[6]

Another challenge in the implementation of the digital twin in some application areas, is the data feedback and the associated direct manipulation of the real object by the digital twin. Currently, the possibilities of technical application are still limited in practice^[1]. In addition, the quality of the data used regarding to topicality, relevance, accuracy and consistency must also be taken into account. This affects the informative value of the digital twin, which, for example, allows statements to be made about the performance of production.^[22]