News of digital twins lands in the Engineer’s inbox. A finger moves to the delete key. He has just processed a hundred emails clamoring for his attention—meeting requests, the calls for help that await him each morning. No wonder he’s so grumpy.
He pauses. A glimmer of hope emerges. Maybe this it; they have figured it out. He swats the hope away. A moment of weakness. It’s impossible -- nobody can replicate a real-life object. God, if there is one, can make real twins. Programmers? Good luck. The email is deleted. The inbox cleared. It’s still early in the day. Time to get to work
But there is a knock on the door. A bright-eyed intern pops his head in.
“Did you get my email?”
The Engineer gets a lot of email.
"Could you be more specific?" he says.
He looks away from the screen slowly, trying to remember the intern’s name. “The one about the digital twin,” persists the intern. He’s breathless with excitement. That email is not going to go away easily. Neither is the broader subject of the digital twin. Nor are these interns. Oh, well. It’s too early to crush hopes and dreams. Better to … what’s the word they use? Oh, yes … nurture. Maybe there has been a paradigm shift, a real breakthrough. The Engineer invites the intern to have a seat.
The State of the Art
"Now what's this about the digital twin?" asks the Engineer. He hopes the ensuing dialogue may impart some healthy skepticism where absolutely none presently exists. Or at least keep people from barging into his office with the latest marketing puffery. “A digital twin is a computer-generated model of a real object,” says the eager intern.
Very good. Actually, a digital twin is only a notion. It is a notion based on the belief that an object in the real world can be completely modeled with computer data. We are not there yet—not even close. A digital twin—similar to its counterpart in real life, a twin in all ways—is impossible. Let me tell you why.
The Engineer wishes he could cue the theme from Mission Impossible.
But first, you should realize that when a company’s entire raison d'être is digital modeling of one type or another—like the email you sent me—you must dismiss their claims of a digital twin as hubris. A real twin, the human kind, if it is an identical twin, is virtually indistinguishable from its sibling. And that’s at birth. After that point, it gets even harder.
“But surely, there are aspects of digital duplicity that are accurate?” insists the intern, high on technology and the salvation it offers.
Possibly. This isn’t going to be as easy as the Engineer had thought. He’d have to dig deeper. Maybe unleash a rant. He wants to tell the intern how we have been making digital representations of objects for a long time, Son. For over 30 years that I’ve been in the business. Longer than you’ve been alive. Too much? He edits himself. It comes out as:
They have been making digital models for almost as long as there have been computers.
Though the digital representations have been getting better and more realistic, to date they fall short of real twin status. A CAD model of a part is an example. But a CAD model is about as similar to the part as a shadow is to the person who casts it. A CAD model may capture the shape of the part—possibly its appearance and a few of its physical attributes, such as mass. But it falls far short of the real thing.
Most obviously … and here he pauses for effect. It’s only in the computer. It’s digital. You cannot hold it in your hand.
“Holding it in your hand is overrated,” dismisses the intern, never before so bold. “If it looks like the real part, and acts like it, isn’t that close enough?”
Brushing aside the impertinence of youth, the Engineer resorts to a favorite parable
The Elephant in the Room
Like the proverbial blind man describing the elephant from having touched it, a claim to have completely defined the part, job done, is an overstatement. Like the blind man, it may be complete as far as the limits of his perception, but for others, the part will require more information. What of its cost, strength, fatigue characteristics, UV sensitivity … other engineering and design factors. Downstream there are manufacturing factors. How the part takes shape from the metal ingot requires a CAM representation. Nontechnical factors include biodegradability, ergonomics, and marketability, among others.
Throw in time, and it really gets complicated. How does it wear, abrade? A part in the field subjected to long-term exposure to severe conditions or ordinary use can look far different than it does when it leaves the factory. Consider a brake pad wearing thin after repeated use. For an architect, a building information modeling (BIM) version of a building is sufficient. Again, one job done. But not for everyone. For the potential tenant, it’s not enough. How will he feel in it? What view will he have? The civil engineer will need to know its seismic survivability, and maybe its rate of subsidence. Each demand adds another layer of data.
And Then There’s Time
A biological twin will be nurtured in the womb, and then will be born, grow, change, age and die. The first question a doctor asks of a patient is how old they are. The doctor knows that your mind and body are different depending on your age. In order for a digital twin to behave similarly to its real-life twin, it too must take time into consideration. Current digital models are mostly time independent. With the exception of motion assemblies (kinematic behavior), digital models of objects show no difference over time. In the current state of the art, a part is “born” after its initial design. It won’t rust. It won’t fail in fatigue. A building won’t subside. A road won’t become potholed.
The so-called digital twins of today make no provision for individuality, either. A digital design, like a CAD model, is but a template for the real object or objects in the real world. Even mass produced objects are can be different from each other due to tolerances allowed, wear in the cutting tool, manufacturing variations, or from the effect of their environment over time. A bullet that is mass produced is made unique in the act of firing it and can be matched to the gun that fired it. In living things, like fellow humans, the uniqueness is understood and apparent.
The intern is getting fidgety. The Engineer may have strayed, broadened the subject, and left the intern gasping for relevance. Time to bring it back to Earth.
Do you have a dog?
“We had pugs growing up,” says the intern.
Silliest of all dogs, the Engineer thinks. But, after editing his thought, it comes out as:
Right, so you can tell one pug from another. Uniqueness and time effects must be considered in in living things. Digital twins, as they exist, even in medicine, ignore those realities. Not because they want to, but because they have to. There’s simply not enough computing power and data storage. Maybe one day will render a future where digital twins can also take into account individual variations as well as changes over time….
A little too philosophical, perhaps. More press releases are calling. There are many more breakthroughs to attend to. But the Engineer is on a roll.
But medicine alone may be the single best use of digital twinnage. The Engineer allows himself a hopeful thought.
Digital Twin of Yourself
Let’s hope this doesn’t happen to you. You go to the doctor. The have a dreaded disease. The good news is that there is a cure. However, it has only been tested in mice. We have no idea how you might react to it.
Wouldn’t it be nice to have a digital twin of yourself who could take the medicine to learn whether you would experience any serious side effects? The digital twin would have your exact makeup, all determined by an examination of your body and your organs. Not the body of some average human, whatever that is. Not of mice. And you would not have to wait years to learn the results. The digital twin could fast forward through a lifetime of taking the medicine in time for your next doctor’s visit.
Seeing It All
Similar to the blind men and the elephant, many stakeholders have modeled an aspect of a real object or process. It is limited to each perspective. It may serve only one purpose.
For example, a rendered image of a part completely describes how a perfect part will look in a perfect world. Of course, it is incomplete and has built-in limitations, an inherent and unavoidable incompleteness. But what if more stakeholders have also made a digital representation of the part for their own purposes? An analyst will add mass and material properties so it can serve as a digital model for stress analysis.
The service department may rig it with sensors and have the part transmit data from the field to help with maintenance, adding the dimension of time to the digital twin. With each additional layer of data, more aspects of the real object are being captured. To someone with oversight, who is able to integrate these aspects, the digital model is filling in and fleshing out. With enough of them, the model may start approaching a digital twin.
A complete digital twin, again, is a notion, an ideal, a goal. We’re a long way off from that. It won’t be until we can model each subatomic particle and predict its interaction with the next—on an individual basis, over time, in an environment from local to cosmic, that a true digital twin can be born.
The Engineer pauses to let his words sink in, giving the intern time to consider the absurdity of the digital twin and the gall of those claims to have produced one.
But Earthly matters are more consuming for the intern. There is a quota of news items to process. The other interns are waiting to go out to lunch.
“Understood, sir,” says the intern, edging toward the door. “We must keep these lofty claims, this hyperbole in check. Thanks for providing a perspective.”
The Engineer looks down at the screen, a dozen more emails have arrived, including two requests for meetings. That’s plenty of time for an intern. But hopefully that little talk will help reduce some of the pop-ins.