Digital Twins, or how to shape reality

It would be great to be able to anticipate or even avoid anything that was about to happen, right? This is what the digital twins promise (and, increasingly, deliver); A vibrant and innovative field of technological development - and seen as an exponent of current trends in hyper-automation of processes - whose global market, according to Gartner, is expected to generate $183 billion by 2031.

A predictable copy of reality

Digital twins are basically virtual ecosystems (or rather copies) that allow the behaviour of objects or processes in the physical or real world to be simulated. A digital twin is a system designed and programmed in such a way that, if it receives the same inputs and under theoretically similar conditions (we are talking about data with exactly known magnitudes) as the physical object or process of which it is twin, it would produce the same outputs as that object or process, so that we could predict with reasonable accuracy what its behaviour would be in a real world.

Thus, we could anticipate the behaviour of a gas turbine if temperature or pressure changes, of an aircraft if the pilots' behaviour patterns change, of a weapon system in the face of climatic changes... or of human hearts or lungs; even of entire patients when attacked by a scalpel or a micro-organism.

Digital twins are created by algorithms fed with sensor data connected - unsurprisingly - to the original model, and are thus able to predict its state and performance in the face of any real-life changes it might undergo. Thus, a digital twin could face all sorts of situations - copying the behaviour of the physical world for us - hundreds or thousands of times - before having to land on it, so that we never have a loose end (or a risk to take) when it comes to putting a chemical process into production, getting an aircraft into the air, or even testing a new medical treatment.

Saving money and risks in a thousand usage scenarios

Clearly, when it comes to testing the (often dangerous) physical limits of a new energy production system, finding out what the efficiency of an expensive machine would be before starting to build and market it, or trying to get early answers to how a particular type of patient would respond to an experimental treatment before clinical trials, digital twins can save time and resources, support the making of, or discarding, costly choices, and of course mitigate risks before moving to the real world (be it the factory, the energy grid or the patient).

But such systems cannot only be conceived as a tool necessarily linked to the introduction of new products, machinery or processes. The maintenance of today's industrial systems is sufficiently complex and costly that many industrial operations managers are beginning to look to them to reduce costs and friction.

Digital twins can thus be a hugely effective way to monitor machinery and make predictive maintenance more efficient, as we can subject the twin in question to all sorts of mistreatment until we can identify its failure point and anticipate costly - or, as we said earlier, dangerous - breakdowns.

Imagine, for example, the case of a nuclear reactor subject to strict safety limits when planning shutdowns or modifying parameters of fuel management, cooling, fission control material layout or the configuration of safety features, to give just a few examples.

But, more than that, the digital twin can be a platform for collaboration and interconnection between teams working together to develop complex systems, providing a framework in which input data and information about their behaviour is shared before it becomes a reality, shortening times and eliminating errors. Or it can serve as a means of defining industrial plant layouts, allowing us to study the movements of people and assets and predict how a change in the layout of a facility might change the efficiency, comfort or safety of operators, for example.

Data, data and more data

It is clear that technologies around digital twins will increasingly become an unquestionable driver of innovation, because they will allow organisations to develop - at increasingly lower costs and in ever shorter timescales - innovative processes or solutions that would probably be expensive and risky to test in situ.

But there is still a long way to go before digital twins are able to deliver complete - i.e. fully predictable - solutions, starting with the fact that in many organisations there is still a huge deficit of the data needed to feed their behavioural models - IoT, sensors, etc. - and thus be able to digitally anticipate reality with a minimum of reliability. That's right; As in almost all areas of digital transformation, here too the quality of information ends up being the greatest constraint and also the greatest challenge.