Deep learning

In 1970, an oxygen tank aboard the Apollo 13 mission malfunctioned, mid-flight. Nasa engineers quickly built 'mirrored systems' to simulate the spacecraft and test their various courses of action, ultimately rescuing the three astronauts aboard the vessel. This life-saving feat of innovation has been described as the inception of 'digital twins' - computerised simulations of physical objects, built using data collected from the objects themselves, that allow their operators to monitor their current state and predict future outcomes. Now, thanks to a combination of cloud computing, artificial intelligence and the internet of things, digital twins are becoming more detailed, more predictive and, as a result, more valuable to a wide range of applications. Already common in manufacturing, digital twins are now being used to simulate and optimise everything from the human heart to the Earth's climate. As a result, adoption is poised to explode. “There is no doubt that the application of digital twins will only increase over the coming years," says Jason Gordon, private sector lead partner for Strategy, Analytics and M&A at Deloitte UK, "and I would fully expect it to increase exponentially.” Technology enablers of digital twins Advances in three areas of technology have made digital twins more viable and valuable. Firstly, cloud computing has made the computing power required to run simulations and forecasts using high volumes of data more widely available. “Our ability to capture and utilise different and increasingly creative sources of data, that is what’s driving all of this," explains Gordon. “The cloud is helpful because of the sort of elastic computing capability that comes along with it.” Secondly, the growing sophistication and falling price of network-connected sensors is making it easier to monitor large and complex objects, in real-time and in granular detail. “Increasingly, the fidelity of digital twins is improving because of a better, more creative capturing of data, some of which will be through IoT sensors,” says Gordon. Read more: Lotus IT director Ivan Sarti on digital twins, recalcitrant engineers and future cars Indeed, online digital reflections are either enabled by or provide the impetus for IoT investments. For companies implementing IoT projects, 18% have already implemented digital twins, and another 70% plan to implement within 12 months, according to a survey by Gartner. And thirdly, artificial intelligence techniques such as machine learning support more sophisticated predictions based on historical data. These predictions will not be infallible, of course, and business leaders will need to understand and reckon with the probabilistic nature of AI-driven prediction. But, as digital reflections grow in life-likeness, they are expected to enable more effective decision making. Advantages of digital twins The defining feature of digital twins, as opposed to traditional simulation, is the use of data collected from and about specific objects in situ. This allows operators to monitor the current state of the object in question, and its environment, and therefore model the likely impact of any changes. Benoit Lheureux, VP analyst at Gartner has explained to Tech Monitor how widespread the use of digital twins already is. "Everyday things around us like elevators, refrigerators, air handlers, industrial pumps, evaporators, generators" are being equipped with sensors and monitored digitally. According to Lheureux, this ubiquity will make elements of public infrastructure more reliable. "There's no good reason for an escalator to fail," he explains. The process of monitoring machines to schedule maintenance to replace a breaking part before it breaks, deemed predictive maintenance, will become more widespread with the expanding use of digital twins. The growing accuracy of digital reflections allows organisations to simulate and experiment with otherwise high-risk operational challenges. One example is air traffic control, says Gordon. “You're not going to experiment with something that is safety-critical like air navigation, so the only way to do that safely is to try things out in a digital environment." Read more: Ocean infrastructure: AI and automation in extreme conditions Another such area is heart disease. In September 2020, researchers at the University of Sheffield announced plans to build the world's first digital twins for the human cardiovascular system. These twins will measure and model the hearts of individual patients, throughout the course of their lives, allowing doctors to simulate and optimise treatments. “A digital twin that works in rea time alongside a patient – changing and ageing with them – will provide a wealth of valuable information to assist doctors in diagnosing heart disease as early as possible," said project co-ordinator Professor Tim Chico, when announcing the initiative. “It may also be able to identify changes that haven’t yet caused any symptoms or signs, providing vital clinical information that can sometimes be missing from a patient’s medical history." Emerging applications for digital twins Not all applications will be so high risk, however. The pharmaceutical industry is using digital twins to optimise drug development, according to a recent report by business intelligence provider GlobalData. The aim is to reduce the number of tests required to validate drug candidates. Governments are building digital twins of the spaces they manage. The Singaporean administration, for example, has built a complete network of digital twins to model the city, which it uses to inform environmental policy and urban planning. David Bicknell, principal analyst for thematic research at GlobalData, believes Singapore is the first of many: "I think that we'll see some innovative examples coming from cities," he says. "London, Paris, Singapore, Tokyo, Helsinki, Seoul: you can imagine these cities will be getting to grips with digital twins." Meanwhile, The UK's Centre for Digital Built Britain is constructing a network of digital twins for infrastructure projects in the country, to improve efficiency, cost-effectiveness and environmental impact. On an even bigger scale, last year the European Commission unveiled plans to construct a collection of digital twins to model Earth's climate. The Destination Earth (DestinE) project will draw on climate and other environmental data, allowing researchers to not only monitor environmental health but forecast the outcomes of climate policy initiatives. From 2025 onwards, the Commission hopes to combine these various systems to construct "a full digital twin of the Earth". As these examples show, the potential applications of digital twins are many and varied. “Digital twins and simulations are being created at all sorts of different levels of specificity, and the most useful ones are being built to solve or to target quite specific business problems," says Gordon. “There is no doubt that business leaders will become increasingly confident with using a digital twin and a simulation and the probabilistic problem-solving that comes along with that, they will get better at asking questions and using the technology in increasingly valuable ways.”

Deep learning

Digital twins set for ‘exponential’ adoption in 2021 and beyond

New global tech hubs are emerging from the Covid crisis

How AI and machine learning can support the renewable energy transition

Forget ‘big data’: It’s the small data that delivers value

Changelog