Do you have a digital twin? Admittedly, the question sounds like something from a bad Sci-Fi movie in which someone has an evil doppelganger who is causing chaos in the virtual world. If you put Hollywood aside, however, the concept of a digital twin is not only real, it is a concept that is rapidly gaining traction.
Having worked in IT for decades, one of the things that I have noticed is that there is a certain evolution to the adoption of technical terms:
- First, someone creates a new buzzword as a way of trying to hype a particular trend or technology.
- Next, a percentage of the IT population begins to use the term, but inevitably some people use the term in a way that is different from the way that it was originally used, resulting in confusion and ambiguity over what the term really means.
- Finally, the IT industry either abandons the term completely or comes to a consensus on its meaning.
Let me give you an example. Several years ago, the term “big data” seemed to spring up out of nowhere. The problem was that at first, nobody could agree on what it meant. An editor at a magazine that I used to write for had asked me to do a story on the so-called big data revolution. I wrote the story from the perspective of finding new uses for business data. Upon submitting the story, I was chastised by the editor who told me that I clearly did not understand big data, because big data only refers to digital video files. Eventually, of course, the term became better defined, and today, use of the term is starting to fade.
Right now, the term “digital twin” seems to be in its confusion phase. The term means different things to different people. Of course, I have my own ideas about what the term means, and would actually like to see the term digital twin split into two different terms–“digital twin” and “intelligent digital twin.” In fact, that is the approach that I am going to take within this article. So let’s go back to the original question… What is a digital twin?
What is a Digital Twin?
At its simplest, a digital twin is a digital representation of a physical object. Although the term digital twin is somewhat new, the concept of simple digital twins has been around for decades.
I started flying light aircraft when I was fifteen years old. As any pilot will tell you, flying isn’t exactly world’s cheapest thing to do, and at that point in my life, my income consisted of whatever I could make from bagging groceries and mowing lawns on the weekend. Because those types of jobs weren’t conducive to paying for a lot of aviation fuel and aircraft rentals, I used to practice landings, navigation, and other skills by using Microsoft Flight Simulator.
Microsoft Flight Simulator was one of the first PC based flight simulations to attempt to accurately model the real world. Granted, PCs of the time had nowhere near the graphical capabilities that they do today, but the software at least attempted to model a few well-known areas. When flying in Chicago, for example, there was a digital representation of what was then known as the Sears Tower. This is an example of a simple digital twin.
The model of the Sears Tower in Microsoft Flight Simulator was a digital representation of a physical object (a building). I consider it to be a very simple digital twin, because it did not actually do anything. The model was useful nonetheless, because it could help you to determine your position when doing visual (non-instrument) maneuvers.
Sticking with the Flight Simulator theme, let’s fast forward a couple of decades. Last week, I was at Embry-Riddle Aeronautical University in Florida going through pilot training on the university’s suborbital spacecraft simulator. In some ways, the simulator could be thought of as a different type of digital twin, although this is the point at which the definition of a “digital twin” gets a little bit hazy.
As you will recall, my original definition of a digital twin was a digital representation of a physical object. The problem with applying the definition to the spacecraft simulator is that like the real space vehicle, the simulator is a physical object. It is a high-fidelity simulator that appears to be constructed from metal and fiberglass. Even so, I consider the simulator to be a pseudo-digital twin because it uses digital technology to provide the illusion of flight. I practiced landing at the Space Shuttle Landing Facility at the Kennedy Space Center (while wearing a spacesuit no less), even though the simulator never moved an inch.
The reason why I chose to use the suborbital spaceflight simulator as an example of a digital twin is because it does something very important beyond allowing for basic pilot training. The simulator is a (partially) digital model of a space vehicle that has not yet flown in space. As such, the simulator is an attempt at accurately predicting the launch vehicle’s flight characteristics. Predictive capabilities are key in the creation of an intelligent digital twin.
An intelligent digital twin is a digital twin that you can learn from. Think of an intelligent digital twin as being a simulation, although a digital twin does not necessarily have to simulate flight. The digital twin revolution that is currently taking hold is more often based on the simulation of industrial equipment, computing environments, or even business processes.
The thing that I find the most interesting about intelligent digital twins is that they can be closely tied to the Internet of Things (IoT) and to Artificial Intelligence. Here is an example that is loosely based around things that several large companies are actually doing:
Imagine for a moment that a manufacturing facility makes extensive use of IoT enabled sensors that monitor the manufacturing process. At the time that these sensors were installed, they were probably put into place as a mechanism for detecting problems with an automated manufacturing process,and alerting the appropriate person to the problem. Because the sensors are digital, however, they are constantly producing data. That data might reflect the health of the sensor, or it might reflect the state of the manufacturing process.
Because the IoT sensor receives sensory input and outputs data that is presumably based on the input received, it is relatively easy to create a computer simulation of the IoT sensor. With that in mind, imagine that the facility has hundreds of similar sensors that all serve specific purposes. All of these sensors can be simulated, and the simulated sensors can be tied together to simulate the entire manufacturing operation. Because real-world sensor data has been collected, nominal operations can be modeled very accurately.
The flip side to this is that because the simulated environment is not real (although it could conceivably be made to reflect the current state of the real sensors), engineers are free to experiment with the simulated environment. In doing so, an engineer might look for ways to make the manufacturing process more efficient, or perhaps find ways to spot problems before they happen based on signals given off by the sensors.
Digital twins have been around forever in one form or another. What makes the current digital twin revolution so compelling is the potential to create intelligent digital twins that can act as predictive analogs to the real world.