Digital twinning: a look at the technology landscape
Digital twins are changing how systems are designed and operated. Understanding the concepts and the enabling technologies is crucial to successfully incorporating digital twins into product development.
The digital twin (DT) represents a manufacturing paradigm shift that is long in the making. The fundamental premise is that for every physical product, there is a virtual counterpart that can perfectly mimic the physical attributes and dynamic performance of its physical twin. The DT exists in a simulated environment, controllable in very exact ways that are not easily duplicable in the real world — eg, speeding up time so that years of use can be simulated in a fraction of a second.
Thanks to the expansion of companion technologies such as artificial intelligence (AI), ubiquitous wireless internet access and inexpensive sensor platforms, DTs are quickly becoming a feasible reality for many companies looking to make better products and more informed business decisions.
Despite all the hype surrounding DTs, the actual concept is rather straightforward. With roots in modelling and simulation, advances in companion technologies, digital thread (to tie it all together) and machine learning (to make sense of it all), digital twinning is on the verge of shifting the landscape of engineering design.
Roots in modelling and simulation
DTs can trace their roots back through the history of modelling and simulation:
- A model, not totally unlike a DT, is a physical or mathematical representation of the characteristics and behaviours of an object, system or process. It dictates how the modelled entity reacts to and impacts its environment and other entities.
- A simulation takes a model or set of models and mimics their operations over time by interjecting artificial inputs (or accepting inputs from a human or an instrumented test object) and monitoring the outputs: this a concept known as ‘live, virtual and constructive (LVC)’ simulation.
Even before computers made it possible to represent tangible objects virtually in software, physical mock-ups representing production systems were sometimes used to understand complex systems better: a great example of this is the full-scale simulators built by NASA to train astronauts to occupy various spacecraft. In recounting the events of Apollo 13’s near disaster, the 1995 movie, named after the spacecraft, gives viewers an excellent basis for understanding the use cases for DTs. In the film, astronaut Ken Mattingly spends countless hours in a functional, exact replica of the ill-fated spacecraft. His goal was to solve the various technical hurdles that were thwarting the safe return of his fellow astronauts. The cost and effort of building exact duplications of the Apollo spacecraft was justified by the enormity of the undertaking and by the amount of planning and practice each mission required.
But what if the cost and effort to create a functional facsimile of a complex system could be made trivial compared to their total lifecycle costs and/or savings? What if, unlike a traditional model, it is possible to use a virtual representation for more than just system design — for instance, using virtual representation to help understand and control supply chain and other business functions associated with product manufacturing? And what if customers could get extremely intelligent predictive maintenance planning based on sharing operational and maintenance data across an entire fleet of systems? This is where the DT could help to change everything about design, construction, operations and maintenance of complex systems. In this context, the Internet of Things (IoT) would be the lifeblood, separating traditional models from next-generation DTs.
Advances in companion technologies
At the heart of digital twinning is a key concept: the virtual and the physical are inextricably linked. Thus, IoT and the more manufacturing focused Industrial Internet of Things (IIoT) have become key enablers that allow data to flow between the digital and physical twins. Embedded sensors on a physical object can monitor all aspects of the object’s operations as well as the operating environment. This valuable data will feed to the object’s DT via IoT for operators and engineers to understand better how a system is operating in real-world conditions.
Reliably enabling a system’s teleoperation requires near ubiquitous internet access. The forthcoming rollout of fifth-generation wireless networks (5G) will bring many advantages to the wireless market that will be necessary for further proliferation of IoT and IIoT. The advantages include increased reliability, more concurrent users and greater built-in support for device-to-device communications. A parallel development, multi-access edge computing (MEC), will help ensure network throughput by offloading cloud processing and maintaining it closer to the sensor nodes, which are foundational to IoT. In short, the processing horsepower packed into today’s inexpensive embedded systems eliminates the need for raw data transport across networks (and/or the internet) to activate processing by high-powered servers.
The digital thread
A fully effective DT needs a closed data loop, or digital thread, that flows from conceptual design all the way to real-world feedback from fielded systems. Embedded electronic products require multiple disciplines to come together to design and manufacture a finished product. Computer-aided design/engineering (CAD/CAE) software suites enable designers and engineers to build the enclosure and mechanical aspects of a product. Electronic design automation (EDA) applications enable schematic capture and circuit board layouts. Computer-aided manufacturing (CAM) software translates the designs into instructions that manufacturing machinery understands to turn the digital into the physical. Each step along the process adds more data to the DT.
The digital thread is the connective tissue that enables the otherwise disparate applications to communicate. Permitting disparate software applications to interoperate, an emerging class of software known as robotic process automation (RPA) enables easily built digital threads. Running at a human user interface (UI) level, RPAs empower disparate applications to interoperate, without expensive software rewrites for each individual application. This capability will prove to be very useful as the digital thread continues to collect data and provide information to the DT from various business systems, such as customer relationship management and supply chain applications. Even after a product has been sold and is in use, the digital thread continues to feed telemetry data to the manufacturer for model refinements on the basis of how a product is actually performing in real-world conditions.
Machine learning turns data into information
All the data moving along the digital thread are impossible for humans to efficiently process on their own. Machine learning technology will be essential to sift through the mountains of data that feed back from fielded systems. Finding anomalies or trends will allow engineers and designers to refine future product iterations in a more predictive fashion than possible today. Cognitive digital twins, powered by AI technology, will allow products to improve over time without any human intervention. In short, instead of just performing mathematical analyses on raw data, a cognitive digital twin would be able to learn, reason, adapt its logic and make informed decisions on its own. The result: the ultimate in technology self-help! The implications of a more cost-effective, adaptable and intelligent product development lifecycle would seem to make any investment in this technology well worth it.
With DTs, every physical product can have a virtual counterpart that can perfectly mimic the physical attributes and dynamic performance of its physical twin. DTs are quickly becoming a feasible reality for many companies looking to make better products and more informed business decisions. Rooted in modelling and simulation, DT has gained traction due to advances in companion technologies, like wireless communications, sensors, AI, machine learning and more. Digital twinning may indeed shift the landscape of engineering design as we know it.
Digital twins vs simulations
A false assumption suggests that DTs are just another type of modelling and simulation. If this were the case, DTs wouldn’t be useful for electronics engineers. Electronic design automation (EDA) software, which enables circuit capture and simulation, has been around for decades. However, ‘twin’ is the emphasis here. It implies the existence is a physical duplicate: of course, under the consideration that the product doesn’t solely live as 1s and 0s in a computer. For product developers who choose to embrace DTs in their design process, this means physical prototypes become even more important. Simulations are only as good as the assumptions made by the person who is running the simulation. However, DTs rely on aggregated real-time feedback from all prototypes being used in various real-world settings. This differentiating philosophy has significant implications for hardware designers.
Testing 5G components and devices with over-the-air methods instead of the cabled methods...
How do you protect yourself from disruption? How do you innovate without radically increasing the...
The media tends to focus on the consumer Internet of Things, but thinking of a test system as an...