How digital twins ground the IoT in reality

Distributed computing and system modeling often involve a degree of abstraction that defies easy use. The ties between real-world assets and their data-based analogs can be frustratingly elusive. Many representations come with inherent disconnects that make it tough to apply their insights.

But what if the data analogs were more closely linked to the source? Digital twins make the internet of things (IoT) a bit more practical by using real-time embedded sensor data to further decision-making.

The concept of digital twinning is almost exactly how it sounds: A twin is a virtual clone of a real-world system or object — but there’s a key distinction. Simulations and system modeling are nothing new. Digital twins stand out by incorporating on-the-fly data into the process. These representations integrate closely with IoT sensors to update their underlying parameters as events develop.

Why is this such a huge leap forward? In short, there’s no such thing as a perfect numerical model, but digital twins track how systems evolve in real life, self-correcting to produce accurate results.

Engineering problem-solvers

IoT imbues digital twins with enhanced dimensionality that benefits many industries. People leverage twinning to simulate factories and extend asset lifetimes.1 Urban planners model cities, construction, and population factors.2 This versatility is the ideal answer to many engineering problems.

Digital twin models work well with machine learning techniques and endpoint intelligence. For instance, dynamic programming and reinforcement techniques are geared for the same type of mathematical optimization that drives twinning.

Source: Getty

Many AI applications are solutions in search of a problem. People know AI is powerful, but as with any nascent technology, they don’t always have the facts to support exploration. With digital twins, the two domains naturally overlap, making it easier to justify computing expenditures.

Digital twins are helpful for problems that benefit from multiple-time-scale analysis. For instance, many dynamic physical systems exhibit inconsistent responses to stimuli over different time frames. These responses can be difficult to quantify, but research has shown that twins powered by sensor data and machine learning adapt to variability well, producing robust, accurate results.3

Reducing exploratory risk

IoT’s proliferation also boosts risks. More devices mean more attack surfaces and increased cybersecurity workloads for engineers. Digital twins aren’t just for simulating systems like manufacturing lines or wind turbines. They also improve operational awareness of the IT frameworks they run on. By integrating networking, performance, and OS data into a model, engineers can try new things without losing sight of the security side effects.

The risk reduction benefits also extend to physical-world hazards. For instance, one study used digital twins to model aerial firefighting operations, which often involve mixed aircraft fleets and dangerous maneuvering.4

Boosting experimental efficiency

Virtual infrastructure also lends itself to practical cost-benefit analysis and experimentation. Siemens is a great example of how this plays out in the real world. In 2018, the company awarded the non-profit BRIDG in-kind grants worth $30 million to explore digital twinning to reduce semiconductor development lifecycles.5 The investment makes sense, considering that machine-learning-aided twins can be millions of times faster than physical modeling.6

Digital twins allow for accurate experimentation even when the actual infrastructure doesn’t. They make it easier to discover behavioral patterns and analyze changes predictively, even in the face of changing conditions.

Choosing IoT technology

Digital twins fall into a few categories, depending on who you ask. While some observers break things down using a product-production-performance scheme, others stick to a component-asset-system-process model or other taxonomy.

The unifying theme behind most classification systems is that they reflect distinct magnification levels. For instance, a specific model might analyze a product or a process. The optimal amount of granularity boils down to the user’s intentions and the type of IoT technology producing the data.

Not every connected device produces viable — or valuable — digital twin data. Compatible IoT hardware must include some form of high-fidelity embedded sensors. This goes beyond installing a few calibrated feedback devices in a smart building or a vehicle fleet. IoT sensors for twinning should satisfy stringent performance standards: low power consumption, ruggedness, and compatibility with secure communication protocols are the bare minimum.

Major twinning players

Modern digital twin implementations mirror the diversity of IoT networks. Sensors that use SPI, Bluetooth LE, and other common protocols aren’t always interchangeable, but they’re compatible enough that engineers can often mix and match software and hardware.

With a bit of forward-thinking system design, there’s no shortage of backends and frameworks at your beck and call. The good news is that there are just as many vendors to match. Players like IBM, Amazon/AWS, and Microsoft/Azure provide customizable frameworks geared toward enterprise use cases. As cloud-native computing giants, they also accommodate integration with existing networks.

Some engineers may prefer a less generalized approach, but this isn’t a huge hurdle. There are quite a few startups focused on specific sectors, like cmBuilder.io, SpaceIQ, and others in building information modeling and AIBODY in the physiology-as-a-service niche — human digital twinning.

No matter the market, there’s likely an existing digital twin framework. If not, engineers can always dig down to the hardware level. Offerings from vendors like Siemens, Bosch, GE, and Schneider Electric make it possible to implement a model in-house.

Future of digital twinning

The digital twin market is hot on the heels of the IoT surge. According to IoT Analytics, the sector’s CAGR could hit 30 percent from 2023–2027, and almost 30 percent of global manufacturers had already implemented digital twin strategies by late 2023.7 The field may still be young, but the value is there.

Twinning stands to redefine not only how engineers explore new ideas but also how they validate ongoing projects. No simulation is perfect, but building digital twins on a foundation of dependable sensor networks can make IoT models far more useful.

————————————-

Ambiq is an Austin TX-based developer of ultra-low power SoCs for IoT endpoint devices that demand complex operations and longer battery life.

https://ambiq.com/