The origins of the digital twin concept can be traced to NASA’s Apollo project, when a capsule identical to the one in space was kept on Earth to simulate the latter’s behaviour
This anticipated the digital twin notion of having one object simulate the effects of another. Yet the space capsule on Earth was not a digital representation but a physical one.
As of now, there is no standard definition of the concept “digital twin”. In Lewis Carroll’s Alice’s Adventures through the Looking Glass, Humpty Dumpty says: “When I use a word, it means whatever I want it to mean”. In the case of digital twin, the various definitions reflect specific characteristics of the technology’s multiple use cases. Given the rapid evolution of technology, the understanding of the term digital twin has undergone almost constant change. The concept has been blurred and has spawned numerous related terms, such as the Digital Shadow or the Digital Angel, thus hindering precise usage of the term digital twin.
Even though there is no common definition for digital twin, different versions agree in some common features, such as the integration of various data sources to permit a digital representation of the physical object or process over its entire lifecycle, based on which various analyses and simulations can be carried out. Simply stated, digital twins are digital representations of physical assets, and, as such, they permit us to predict the assets’ future.
Crystal ball or magic mirror?
Given today’s data volume and computational power, it is no longer unreasonable to imagine a computing system that tries to simulate the actions and interactions of different assets. In such a digital mirror world, our virtual machines might be even equipped with cognitive abilities and decision-making capacities. If we fed this virtual world with our own real data, how similar to us would they behave? Would it eventually be possible to create a digital copy of our assets, a virtual reality as realistic as life itself?
The crystal ball analogy is an interesting one, but it falls short in one area: the crystal ball/digital twin will work only if we continuously feed it data from the real world. In other words, the crystal ball needs connection with the real world to make accurate predictions. So at this point, we have to change our analogy. The digital twin is not a crystal ball after all but is acting like the mirror of Snow White´s evil stepmother. The stepmother possesses a magic mirror, which she asks every morning, "Magic mirror on the wall, who is the fairest one of all?" The mirror always tells her that she is the fairest. She is pleased because the magic mirror never lies. But when Snow White is seven years old, her fairness surpasses that of her stepmother. When the queen asks her mirror her daily question, it tells her Snow White is the fairest. In other words, new and changing data have led to new conclusions.
This aspect of the digital twin is extremely relevant, since continuous data feeding and contextual awareness are required for prediction. The digital twin cannot forecast anything when it is disconnected from reality or when it lacks data about actors, actions, and contexts.
The Picture of Dorian Gray
From an engineering point of view (not a fairy tale point of view), the correct definition of a digital twin is an integrated multi-physics and probabilistic simulation of a system that uses the best available physical models and data-driven techniques to mirror the life of its corresponding physical twin. This entity will mimic reality from information input from the physical world and ideally will create a mirrored world.
The more data the better to get an ultra-realistic presentation and make useful predictions because the digital twin is not a one-time shot. The use of evolving data means a key strength of the digital twin approach is its ability to provide an accurate description of objects that change over time. A validated model can provide a snapshot of the behaviour of an object at a specific moment, but using that model within a digital twin can extend the use of that model to timescales over which the object and its behaviour will change significantly.
This provides a new dimension for digital twin – the lifecycle view, wherein the digital twin is a replica which grows parallel to the growth in the real entity. Once we add this lifecycle approach, we immediately think of the short story “The Picture of Dorian Gray” by Oscar Wilde. The main character sells his soul to the devil in exchange for never growing old. Instead of the passage of time taking its toll on his physical body, the marks of age and his sins accumulate to horribly disfigure a portrait of the youthful Dorian hidden in the basement of his London home. This is the most sought-after feature of digital twin usage: the physical entity will be as good as new, while ageing will be visible in the digital representation.
As long as the digital replica displays all potential hazards, giving us time to mitigate risks, the physical entity will be “immortal”. Indeed, Dorian Gray only dies when he destroys the picture; with this action, the ravages of time and his hedonistic life are transferred back to him, and the original image in the picture is restored. A hallmark of engineering is that we want assets that do not age. Immortality should be assured by the combination of maintenance and preservation actions.
The digital twin as a virtual information construct that fully describes a physical product from actual sensor data, current data, past actual data, and future predicted data will be useful for a variety of purposes. Although the prospect of building a digital crystal ball, magic mirror, or evolving portrait might sound ominous, we should carefully discuss it. The potential benefits are obvious; many big problems might be solved with one or all of these predictive capabilities, including potential harmful scenarios with serious consequences for our assets. This is the essence of Dorian Gray standing before his portrait, considering both himself and the differences from himself, and gazing into the portrait “as if he had been looking into a mirror after he had done some dreadful thing”.
Alice’s Adventures Through the Looking Glass
So far, we have conceptualised a unidirectional digital twin and exemplified it as Dorian´s picture where Dorian is the passive observer of ageing or other consequences. However, the mechanism linking reality with the digital entity is bidirectional. The digital twin may eventually affect the real entity, modifying the course of its history by entering one of multiple virtual spaces depicting a single real space. Simply stated, the digital twin is a simulation engine where alternate ideas or designs can be explored and ultimately influence the reality.
Coming back to Lewis Carroll – in Alice Through the Looking Glass, Alice is taken into an amazing world when she climbs into the looking glass. Much like Alice, we use existing technology to step into an amazing new world generated by our physical models and artificial intelligence, recreating potential futures for our assets. Yet we need to reassess this. Alice finds herself climbing into the looking glass and wandering through an amazing new world. The problem is that, as in Alice’s dream, it is getting more difficult to distinguish what is real and what is not. In fact, we are not even sure if we can say digital is different from reality, as the assets and the replica act in a coordinated and seamless way.
This bidirectional data flow implies an active role by the digital twin. This, in turn, enables real-time monitoring of systems and processes and timely analysis of data to prevent problems before they occur, schedule maintenance, or modify operational profiles. When Alice climbs through a mirror and enters a new world, she first thinks this world is very similar to her own, but it turns out to be quite different. The precise set of rules and behaviour that apply in this mirror world are not entirely clear.
If we want to solve our version of Alice´s problem – when we are confronted by a crazy set of future scenarios and rules, and there is chaos everywhere – we need proper models, good quality data, and direct links to and consequences for our everyday environment.
If we can do so, we will be able to distinguish what is real from what is not.