The idea that a R20;thing” in the real world has a digital representation in cyberspace may have its roots in William Gibson’s 1984 novel Neuromancer, but when companies talk about “digital twins” these days, they mean something considerably more recent, just from the past couple of years. In this instance, the process of designing products digitally and expressly tying the manufactured product to its design representation over a period of use is something that Industrie 4.0 is only just beginning to sort out.
Their appearance is clear, as a digital twin technology approach can facilitate:
An aircraft manufacturing plant uses digital twin technology from Swim.AI, for example, to monitor, analyze and optimize inventory management using data from tags on various components.
Xcel Energy Inc., a major U.S.-based electric and natural gas company, implemented a digital twin app using GE’s Predix Platform to help increase the efficiency and tracking of information across its hundreds of employees. This reduces the time spent manually updating data and cases, resulting in enhanced visibility into asset and plant performance, as well as easy visibility into best practices across the enterprise.
Digital twins improve development by allowing developers to use programming instructions to directly manipulate the abstract version of the device itself, letting the system underneath the twin concept sort out the niceties of transferring new data and machine state to the physical twin. The immediate focus is thus on the application logic rather than the details of communicating with the device. For example, Amazon Web Service’s Greengrass technology allows developers to code against a device shadow.
With digital twins, IoT device PLM can be improved by aggregating IoT data with data from CRM, ERP and product specifications. This makes it easier for product managers and engineers to correlate design, customer satisfaction, device performance and device reliability.
IoT brings more comprehensive modelling
The idea of virtual representation of the real world is as old as computing itself. However, early models focused on specific problems tailored to industrial assets. “This narrow focus tends to limit solutions using those models to particular areas and a narrow class of assets, reducing the applicability of these models and the ability to deliver differentiated customer value through true end-to-end system or network optimization,” said Patric McElroy, vice president, chief software engineer at GE Power.
For example, in the power industry, the focus was traditionally on a part of the network rather than the entire system itself. Modern digital twin technology that brings in data from across different systems can help engineers achieve a more diverse set of goals. They can also make it easier to use machine learning to correlate a much wider set of data to look at relationships across assets to enable and deliver true network-level optimization. Cross-referencing data also makes it easier to detect and correct anomalies.
Digital twins can make it easier to correlate data from multiple IoT devices for industrial IoT. Enterprises can equip machines with sensors that collect operational data that reflects the machines in the context of the environment as a whole. “Using a digital twin, you can consolidate and analyze these data sets, as well as replicate production processes in the virtual world,” said Gerald Glocker, director and chief product owner for engineering cloud services at Bosch Software Innovations, GmbH. Over time, deviations in performance might become apparent. Manufacturers can then take action and optimize their production processes.
Digital twin technology can also be of use in the context of a connected building. For example, designers could simulate how a building is used based on historical or comparative data and test out changes in the building’s design. This might call attention to rooms that are wasting energy or are used rarely, Glocker said.
Digital twins needn’t be restricted to designed, manufacturable aspects of production. Twin modelling is also being used to create interactive models of the Earth to help enterprises make better predictions. Descartes Labs Inc. is building such a model by fusing data from satellites, logistics data from IoT devices on trucks and other sources to correlate commodity production with enterprise processes. Fritz Schlereth, Descartes Labs’ head of product, said, “By simulating what is happening on the planet, you can begin to predict production and demand as commodities are generated, predict substitutions within the supply chain and predict shifts between supplier and consumer both globally and locally.”
Useful models need to allow developers to take advantage of the global scale of the digital twin, must be accessible via the same API and have to consider historical data. The models need to run continuously onward to offer new insights and intelligence. Schlereth said, “The key to making the digital twin accessible to users is to clean the data so it is ready for scientific use and to index the data so that it is searchable. Making the data searchable requires storing metadata that describes each image and indexing it efficiently.
Digital twins can also provide an abstraction layer for developers to make it easier to securely code IoT apps. Ian Skerrett, an independent IoT market adviser, said, “The importance of a digital twin is that it allows integration of the physical device to other OT/IT systems. An advanced API will make it easier for developers to do this type of integration.”
For example, Bosch IoT Things enables applications to manage digital twins of their IoT device assets. Applications can store and update the data, properties and relationships of their domain’s assets and get notified of all relevant changes via well-defined APIs. Those APIs are exposed via HTTP, WebSocket and AMQP.
This kind of an approach makes it easier to manage and organize digital twins of IoT devices, find things by their dynamic properties or static attributes, and securely communicate with them. Bosch has aligned these capabilities within the open source Eclipse IoT framework, Eclipse Ditto. For developers, this means that APIs of Eclipse Ditto are fully compatible with APIs of Bosch IoT Things. This brings a lot of flexibility, sustainability and trust.
Skerrett said enterprises should explore the openness of their digital twin technology in order to cultivate long-term value. He explained, “We are seeing a lot of the existing industrial vendors that have simulation capabilities extending these to be called digital twins. The next step needs to be better developer tools to create digital twins independent of engineering software, like computer-aided design systems. If digital twins are to become a general industry concept, there will also need to be some work done on consistent definitions of capabilities and APIs.”
Product lifecycle management
Another use case for digital twin technology championed by PTC lies in correlating data related to a device across its entire lifecycle. Francois Lamy, vice president of PLM solutions management at PTC, said, “For us, a digital twin combines data from the device itself along with the experience of the product.” This can include aggregating data from design artifacts, manuals, service and parts information. The data might come from engineering applications, CRM, ERP and service management applications.”
This can help bring new insight to frontline workers, as well as product managers and engineers working on new devices.
For example, a car manufacturer might use IoT data to observe that drivers in Switzerland go through breaks faster because of the hills there. Lamy said, “It is not like this is in and of itself remarkable, but if I can put that information in the context of my design, I can focus on the actual usage of the product.”
From this perspective, the digital twin provides a coherent view of the product. It also requires figuring out how to orchestrate data from these different systems in a way that is useful for engineers, managers and frontline workers in their own context. This involves protocol bridging across different technologies and interfaces. “The big issue is that you are swimming in oceans of data and it is locked in various silos,” Lamy said.
The fields of application for digital twin technology are manifold and not confined to a specific industry or area; they can be used for a wide range of scenarios. On one hand, there are digital twins that just model and represent a single sensor within a device. On the other, there are digital twins that reflect a whole campus of buildings with a lot of aspects regarding energy, usage, topology and more.
So, what’s next?
There is potential for the adoption of digital twins to increase as companies evaluate use cases and find out where they can be beneficial to them. “In the coming years, we expect greater customer requirements in regard to the modeling of twins, which includes semantics and simulation,” Glocker said.