A digital twin is a virtual representation of a real item/system, the physical twin. A digital twin is not a mashup of Arnold Schwarzenegger films. The term digital twin has been rapidly gaining popularity in mechanical design. For one thing, "digital twin" has a lot more cachet than the colloquially used "CAD model." But are they the same thing? A CAD/3D model, at its most basic, is a digital representation of geometry. That's it. So, in its simplest form, it is a digital representation of just one aspect of reality, the shape.
Fabricated representations for the purpose of aiding in understanding the complex is nothing new. The human form, for example, has been copied and replicated for a myriad of reasons from science to entertainment and from the highly realistic to the abstract. A mannequin is one such usage for representing human form. This reductive human exterior is far from an identical doppelganger and has limited utility, aside from helping to sell turtlenecks.
In addition to the looks-like representation, many more layers of analytical representation can be added to the digital twin. These additional layers build on the looks-like with works-like representations. Material properties can be added to create representative mass, mechanical properties and appearances. A crash test dummy's purpose isn't to look appealing but to "work-like" a human in a collision and to do this it needs to approximate the shape, mass distribution and joint movement of a person.
Predicated on the geometry and material properties, constraints and loads can be added to the digital twin for complex, physics-based simulation. Commonly, this is a computational (finite elements) statics analysis. Other analysis types include dynamics, fluid flow and electromagnetism. These analyses must be considered prudently to avoid misleading results. When done wisely, meaning with expertise, caution and corroboration, the insights gleaned can lead to exceptional value delivering decision making. Attempting physical testing for the same results can be prohibitive due to the resources or duration (lifecycle) needed. Analytical computational modeling can require expensive software and know-how and may not fall directly under one master model but as a selective offshoot. There are many paths and approaches for using analytical simulation. There are many variations and purposes for selective human representation like medical simulation, crow deterrence, companionship and animatronic presidents.
Vendor provided CAD files are almost always stripped of material properties and other metadata due to a combination of the nature of neutral file formats and the desire to protect intellectual property. These files contain information for use, but it can take significant effort to manipulate and update the files to in-house standards. Back to the analogy of creating a literal human counterpart, a vendor CAD file is an anatomical arrangement of bones.
At each detailed stage of digital twin development, the fidelity spectrum can range from thes singles to thousands of hours of work. At the most fundamental execution of the geometry stage, perhaps only the exterior surfaces are considered and the internal features and parts aren't represented. Advanced geometric detailing may include details whose utility is more cosmetic than functional, like decals, threads and cable routing.
The classic “step by step pencil sketch of a person" shows discrete presentation levels from connected ovals to photorealism. These levels serve a purpose (beyond helping to guide the drawing illiterate). That purpose is feedback. A crude composition sketch is preliminarily for visualization and consensus on the positioning and framing. It serves to determine the character’s location, angle, movement position and so on prior to detailing. Detailing consumes a disproportionate amount of time and effort. Minimizing the delta on scrapped detailing work is "found money" that can be put towards a project's success. Roughing is strategic zooming out. Detailing is strategic zooming in. A rough outline of a person made of simple shapes captures the story and perspective of the frame. After that, zoom in to the face. Rough in vertical and horizontal lines to position the features on the face. After that, zoom in to the eyes. Zoom out. Zoom in. Zoom out. Zoom in. Zoom win.
So, how much detail is appropriate? Like most things in development, it correlates with the application, release stage, resources, and future capability building. Applied resources should be commensurable with expected value of risk and reward. That's the textbook answer. The answer from the trenches is it's a perpetual series of educated trade-off bets between opportunity cost (scheduled risk) versus cost of poor quality (technical risk).
A rocket ship involves huge financial, political and human life risks with small factors of safety. So, I would hope, for a rocket ship there exists a digital twin database of the highest fidelity at all levels. When designing Unnecessary Inventions like chopstick extensions for your AirPods, a trial-and-error 3D printing approach is probably A-OK.
Premature detail is misallocation
Over detailing ahead of necessitation might be construed as a double positive. Extra work (above and beyond!) and ahead of schedule (proactive!). Work is only truly extra if it is done in addition to all the other necessary work without extra resources. Work is only ahead of schedule if it ever actually needed to be done and was not completed at the cost of putting other work behind schedule. Focusing a resource one way is diverting it in another. Over allocation on one area means under allocation on another area. Continuation of risk reduction in one arena comes at the cost of neglecting another arena's risk. The goal of managing a project should be to systematically build confidence intervals by de-escalating risk in a strategically distributed way. This is the real art and need of management, to gauge risk in a high-level and unbiased way. It's a dangerous gamble to leave one of a project's meta problems disproportionately weighted compared to the rest. Problems with outsized risk are the ones that can torpedo a project.
Under detailed and released is tech debt
Highly detailed geometry, like wire routing splines, can be used for creating wire maps and harness drawings. If updates or configurations are added in the future, the new measurements and documentation can be created lightning fast and almost effortlessly when compared with the alternative of having to do a physical mockup. If the digital twin wire routing is ignored it also increases the potential of an interference or design for manufacture/assembly oversight as it's easy to lose track of all the details on the screen. The fine details that can be created also add that subtle, but very noticeable visual wow factor. The ability to catch errors, adapt easily to revisions and create stunning images virtually is the opposite of tech debt, it's paying it forward.
The key to successful implementation is education. Educate oneself and others on what the practical capabilities are for Computer Aided Design / Engineering. Designers are typically underutilizing software design tools they already own. Simulation tools are powerful but tricky. Laypeople way over index their belief in the power of simulation, but designers are also guilty of this. The best thing is to stay abreast of the tools and grow into them strategically based on an individual's career goals and the company's needs. Be The Thinker, not a dummy.