In recent years, the word Digital Twin has become a popular concept. However, with the continuous interpretation of digital twins by industry and academia, its meaning has become more and more confusing, and the boundaries with other related concepts have become more and more blurred. Many people are puzzled by what digital twins are, what they can do, where the boundary is, and what relationship they have with modeling and simulation. This paper attempts to make a superficial analysis of some of these doubts.
1 What exactly does Digital Twin mean?
Since there is no consensus on the Chinese translation of Digital Twin (this issue will be mentioned later), DT is temporarily used to represent Digital Twin.
In the field of information, it is common for a concept to have multiple interpretations. However, although most concepts have vague meanings and definitions at the beginning, after a period of discussion and precipitation, they will gradually form a consensus, such as cloud computing. DT is very interesting. At the beginning, its meaning was relatively clear, but with the deepening of research, the definition and connotation became more and more vague. In addition, although there are many different definitions and interpretations of many concepts, the general differences lie in either the perspective and emphasis of the problem are different, or the level of detail of the interpretation is different, or the way of writing is different, while the subject of the concept itself is definite. Like DT, different definitions point to different subjects, but they are rare.
The term "DT" is generally believed by the industry to be a concept proposed by Professor Michael Grieves of the University of Michigan in 2002 for product lifecycle management (PLM). At first, it was not called Digital Twin, but Mirrored Space Model (MSM). Later, John Vickers of NASA named it Digital Twin [1]. The attribute of its model is very clear. Although it did not attract much attention at that time, there is no ambiguity. DT is a digital model.
However, as NASA introduced it into the NASA Space Technology Roadmap, the meaning of DT has changed significantly. The explanation given by NASA is as follows: DT makes full use of physical model, sensor update, operation history and other data, integrates multidisciplinary, multi physical quantity, multi-scale, multi probability simulation process, and completes mapping in virtual space, thus reflecting the full life cycle process of corresponding physical equipment [2].
The body of the DT becomes a simulation.
This report also clearly points out that NASA's Digital Twin is "Simulation Based Systems Engineering".
The main body of DT becomes system engineering again.
NASA's interpretation of DT is probably the root cause of the ambiguity of the definition and connotation of subsequent DT. We might as well list some representative definitions.
(1) A DT is a real-time digital copy of a physical device [3].
(2) A DT is a digital copy of a living or nonliving physical entity. By connecting physical and virtual worlds, data can be transmitted seamlessly, so that virtual entities and physical entities can exist simultaneously [4].
(3) DT is a digital representation of an asset, process or system built manually or in a natural environment [5].
(4) DT is a software representation of assets and processes, used to understand, predict and optimize performance to improve business [6].
(5) DT is a virtual representation of the actual product or process, used to understand and predict the performance characteristics of the counterpart [7].
(6) DT is a coupling model of real machines running on the cloud platform, and uses data driven analysis algorithms and integrated knowledge of other available physical knowledge to simulate health conditions [8].
(7) DT is a dynamic virtual representation of a physical object or system in its entire life cycle. It uses real-time data to realize understanding, learning and reasoning [9].
(8) DT uses digital copies of the physical system to perform real-time optimization [10].
(9) DT is a bridge between the real world and the digital virtual world [11].
Figure 1 shows the DT related parts mentioned in the above definitions. Including ① physical object, ② data, ③ model, ④ simulation and ⑤ simulation results.
Figure 1 DT related parts
These definitions point DT to different parts of Figure 1.
Category I: Definition (1)~(5) defines DT as digital copy, digital representation, software representation or virtual representation, pointing to ③, that is, DT is a model that updates with physical objects in real time, because digital copy, digital representation, software representation or virtual representation all belong to the category of models.
The second category: define (6)~(8) to point DT to ③ and ④, that is, DT is model plus simulation.
The third category: definition (9) points DT to ② and ⑤, that is, DT is a bridge connecting physical objects and models.
So, as a rigorous academic term, which one is more reasonable? Let's analyze it.
The second category defines the combination of modeling and simulation as a new concept, which is unnecessary and unreasonable on the one hand. Although model and simulation are closely related, they are two things. Model is a description of things, while simulation is a variety of activities based on the model. As two numbers have different dimensions, they cannot be directly added to one number.
The third category sounds very provocative, but it is the most unreasonable. If the data from physical objects or simulation feedback is called "bridge", it is understandable, but neither of these two types of data can be called Twin of physical objects.
Therefore, relatively speaking, the first definition is the most reasonable, that is, DT is a digital model of physical objects. But this model can receive data from physical objects in real time, so it can evolve continuously to keep consistent with physical objects. Of course, it does not mean that the previous model does not have evolutionary characteristics, but the previous model evolution did not emphasize real-time.
In the understanding of DT, there is another puzzling question: should a DT contain physical objects, namely ① in Figure 1? This also makes many people confused about the relationship between DT and information physical systems.
The source of this problem also comes from the relevant literature of NASA and the US Air Force Research Office [12-13], which believe that the concept of DT consists of three different parts: physical products, digital/virtual products, and the connection between two products. That is ① ② ③ or even ⑤ in Figure 1.
But obviously, there is a logical problem with this explanation. If the DT contains a physical system, then Twin has no reference. Because Twin must be relative to another person (or object). Of course, it is reasonable to call ① and ③ Twins together, that is, twins, but not Digital Twins, because one of them is digital and the other is physical. Therefore, the physical object and the digital model should be distinguished in the DT concept.
According to the previous analysis, if DT is defined as a digital model of a physical object, the relationship between DT and the Cyber Physical System can be easily clarified, that is, digital models, various activities (simulations) based on digital models, physical objects, and connections (data and simulation results) between digital models and physical objects form an information physical system, as shown in Figure 2.
Figure 2 Relationship between DT and information physical system
In addition, there is a question that may cause ambiguity, that is, should DT include the data collected from the physical system, namely ② in Figure 1? In my opinion, DT concept does not need and should not include such data, because the model here is based on the real-time evolution of these data, so the data information will be reflected in the model. There is also a special technology for data, namely Digital Thread technology, which can be used to deal with data problems related to Digital Twin.
Let's take a look at the Chinese translation of DT. At present, although the word "digital twins" is widely used, the terms "digital twins" or "digital twins" are still often used.
DT was first translated into digital twins or digital twins. Later, it was verified by experts to be a digital twin. The general meaning is that twins refer to two people, namely twin brothers or twin sisters. The corresponding English word should be Twins, and Twin refers to one of the twins, so it should be called twin. In fact, the word "twin" is rarely used alone in Chinese. In most cases, it is used together with brothers and sisters, such as twin brothers and sisters. However, in terms of the accuracy of word meaning, it is more reasonable to use digital twins to correspond to Digital Twin than digital twins.
Lenovo's various interpretations of the concept of DT mentioned earlier translate DT into digital twins, which means that the fact that DT refers to a model by default does not include physical systems. Because twin refers to one of twins, it obviously does not include the corresponding physical system itself. If it includes both the model part and the physical part, it is more appropriate to call it digital (digitized) twins, although it is not rigorous, because as mentioned earlier, the physical system is not digital.
In the Chinese literature, there is another interpretation of DT, that is, it is interpreted as a modeling process and method, also known as digital twin technology, and the model constructed by using this technology is called digital twin model, or digital twin. This explanation obviously does not correspond to the basic meaning of DT.
The vague understanding of DT concept is also an important reason why Chinese translation has not been unified.
In conclusion, the author believes that translating DT into digital twins is more in line with its original meaning. Digital twins can be understood as follows:
Digital twins are digital models of physical objects, which can evolve in real time by receiving data from physical objects, so as to be consistent with physical objects in the whole life cycle. Based on digital twins, analysis, prediction, diagnosis, training, etc. (i.e. simulation) can be carried out, and simulation results can be fed back to physical objects to help optimize and make decisions on physical objects. Physical objects, digital twins and simulation and feedback based on digital twins constitute an information physical system. The technology facing the full life cycle of digital twins (construction, evolution, evaluation, management, use) is called Digital Twin Technology.
2 Modeling and simulation technology behind digital twins
As an academic research, we need to adopt a reasonable and unambiguous concept, rather than entangle in NASA's interpretation of DT. However, the real purpose of NASA Digital Twin mentioned by NASA cannot be ignored. In fact, many people's enthusiasm and research motivation for DT comes from the infinite vision of NASA Digital Twin for the future of manufacturing industry.
Although we do not agree with NASA's interpretation of DT, from another perspective, NASA does not actually define DT as a serious academic term. Its real intention is to emphasize the value of simulation, that is, NASA's aircraft development needs to follow the concept of "simulation based system engineering".
Simulation technology almost came into being with the birth of computer technology. Since the 1950s, it has experienced more than 60 years of development. Because simulation is based on the establishment of models, in order to highlight the importance of modeling, modeling and simulation often appear together, that is, Modeling&Simulation, often abbreviated as M&S.
From the following passage, we can see the height of American understanding of modeling and simulation. This passage is excerpted from Resolution 487 adopted by the United States Congress on July 16, 2007 [14].
The United States of America is a great and prosperous country, and modeling and simulation has greatly promoted this great and prosperity.
In the United States, modeling and simulation is a unique application of computer science and mathematics. It is based on the validity, correctness and repeatability of models or simulations, and the ability of thousands of Americans to develop these models in the modeling and simulation industry.
Members of the government, industry and academia in the field of modeling and simulation have made outstanding contributions to the overall welfare of the United States. The following examples can partly reflect the contributions of modeling and simulation to the United States, although these contributions are numerous:
(1) During the Manhattan Plan, the understanding of nuclear fission was expanded through the simulation of the earliest reproduction of nuclear chain reaction process, which ultimately led to the end of the Second World War;
(2) As the basic element of the "inventory management plan", the President of the United States was able to ensure the security and reliability of the nuclear weapons inventory for more than ten years without conducting real nuclear tests, and demonstrated the country's commitment to nuclear non-proliferation.
From the above two contributions, we can see the great power of modeling and simulation. In this resolution, there are 11 such cases. In the resolution, it was also clearly stated that modeling and simulation are "National Critical Technology".
The following content is more impressive.
The Higher Education Act, promulgated in 1965, is the first higher education legislation in American history. By expanding the role of the federal government in the financial field of higher education, it enables the federal government to intervene in the development of American higher education and lays a foundation for the popularization of American higher education. This law is a milestone in the reform of American higher education [15].
The new versin of the Higher Education Law, revised and effective in December 2018, specifically includes modeling and simulation as an important content (20 U.S. Code § 1161v - Modeling and simulation) [16], and uses a lot of space to explain how the government and society should promote the popularization of modeling and simulation technology in university education. In the whole bill, we haven't seen any other technology enjoying such high treatment, even though the AI technology that has been in full swing in recent years hasn't appeared.
It can be seen from these two documents that the foundational and long-term value of modeling and simulation technology has gone far beyond its technical field and will have a significant impact on national interests and national security.
In the manufacturing industry, modeling and simulation has also been playing an irreplaceable role. In 2000, DARPA, the Ministry of Commerce, the Ministry of Energy and NSF jointly released a national manufacturing development strategy research and promotion plan "Integrated Manufacturing Technology Roadmapping (IMTR)". IMTR proposed six "major challenges" facing the manufacturing industry in the future, that is, to become a sophisticated and efficient enterprise, improve the responsiveness to customers, become a fully connected enterprise, maintain environmental sustainability, conduct knowledge management, and be good at applying new technologies. It is not difficult to see that these six challenges are still in place today. IMTR further proposed four types of technical countermeasures to meet these six challenges, namely, manufacturing oriented information system, modeling and simulation technology, manufacturing process and equipment, and enterprise integration. This shows the importance of modeling and simulation technology for manufacturing industry.
In 1997, before the release of this report, the US Department of Defense reformed weapons acquisition. The most important reform was to propose "Simulation Based Acquisition (SBA)", which means that modeling and simulation are applied to the whole life cycle process of weapons from demand analysis to final retirement. The F35 of Lockheed Martin is the first product developed by introducing the SBA concept. In November 2017, it was reported that Lockheed Martin ranked digital twins as the top of the six top technologies in the future national defense and aerospace industry. Behind Loma's digital twin is SBA, that is, simulation based acquisition, which is the same as NASA's simulation based system engineering.
Modeling and simulation technology was born in the United States more than 60 years ago, and has been playing an immeasurable role in the national interests of the United States. However, Americans never hype modeling and simulation. The reason is worth thinking deeply.
3 pairs of digital twins should not be over interpreted and should keep a clear mind
The popularity of the word "digital twin" is largely due to its own communication characteristics. Its image is easy to understand, and even laymen can understand it literally, and it can also trigger unlimited imagination. It is really a good word for science popularization or commercial promotion. But as an academic term, it lacks due preciseness, at least so far. At present, there are two tendencies worth paying attention to. First, the connotation of digital twins is constantly enlarged in concept, giving more and more content, resulting in overlapping or even overlapping with other concepts; The second is to label traditional technology applications as digital twins. These two tendencies are not conducive to the development of digital twin technology.
As mentioned earlier, behind the Digital Twin are modeling and simulation technologies. The most attractive part of digital twins is the combination of digital models and the Internet of Things, and the ultimate purpose of this combination is to polish the model closer to the real system. The Internet of Things technology provides a new and powerful means for modeling. In the absence of sufficient understanding of the mechanism of complex systems, the system can also be modeled using artificial intelligence technology based on the collected data. This is the development and supplement of modeling technology. And model-based analysis, prediction, training and other activities are the things that simulation should do. So digital twin is essentially the application of new generation information technology in modeling and simulation.
After more than half a century of development, modeling and simulation has formed a fairly complete and systematic technical system. In fact, in the field of simulation, the methods and technologies of modeling and simulation using dynamic real-time data have been studied for many years, such as Dynamic Data Driven Simulations (DDDS), embedded simulation, hardware in the loop simulation, etc. Of course, digital twins, as an important part of simulation technology, will further promote the development of modeling and simulation technology through the integration with the new generation of information technology.
In addition, many of the modeling and simulation technology systems can be directly used for the research and application of digital twins, including theories, methods, standards, tools and platforms, and there is no need to repeat development in the name of digital twins. This is also the reason why NASA did not establish a new technology system for Digital Twin itself in the future, although it proposed the goal of achieving Digital Twin by 2027. Because of the required technologies, standards, tools and platforms, most of them already exist in their modeling and simulation technology systems. It is a waste of resources and time to change the label and make a similar set of things.
The popularity of digital twins is very important, although it is important to have a good name, it is more important to benefit from the help of industry giants such as Siemens and GE. The sober insiders know that most of the core technologies related to digital twins are in the hands of foreigners, such as sensing technology, CAD technology, CAE technology, PLM technology, VR/AR technology, etc. Siemens, Predix, Dassault, PTC, SAP, Ansys and other enterprises are exaggerating the fascinating prospect and application value of digital twins because they really master the core technology necessary for digital twins. With the rise of digital twins, the products of these enterprises are really hot. The industry's enthusiasm for new concepts has its own purpose, and insiders are well aware of it.
However, these core technologies and products mentioned above are exactly what China lacks most at present. However, it is interesting to note that in China, the research on digital twins tends to ignore the importance and focus on peripheral and non core things, and the enthusiasm for its concept is far higher than that of the United States and other industrialized countries that put forward the concept. Most of the time, we are keen on the concept of hype, saying more, doing less, doing more false things, and doing less real things. In fact, we do not know that we are cheering for others.
For those of us who are seriously short of core technology support, we should keep a clear head in today's hot words continue to emerge. It is the absolute principle to immerse oneself in the research and development of the key technology of neck clamping. Otherwise, we will waste both resources and time, and ultimately lose great opportunities for development and strategic initiative. (The article is from an AI expert)
