Digital twins: virtual product testing and maintenance

By bridging the gap between the virtual and the real worlds, digital twins are powering a “fourth industrial revolution”. The twins are digital replicas of physical objects (eg, a car) or systems (eg, a production line). They enable virtual modelling to analyse and optimise real-world products and processes. 

Use of digital twins is becoming common in all sectors. In urban planning, they are used to model traffic flows, energy use and carbon emissions. In the automobile industry, they allow researchers to run crash-test simulations countless times without having to destroy cars. They also support sustainability initiatives: for example, using “Virtual Singapore”, a digital twin of the city state, researchers determine where best to build solar panels for power generation by analysing light and temperature variations across different locations.¹

Digital twins can facilitate cost-reduction and operational efficiencies. They can be used to plan maintenance, spot emerging problems, and simulate the effect of upgrades and design changes. By allowing researchers to virtually test multiple versions of new models before producing a physical prototype, the twins can help reduce the incidence of product defects and also time to market.² In manufacturing, they help with training of personnel and production facility planning.³ And, once a product is in use, predictive maintenance helps reduce downtime: the digital twin can be fed with real-time data to monitor and simulate product behaviour in different environments. This can help expand the scope within which the product can operate and also prolong its lifetime.⁴

The versatility of digital twins and their rapid integration across sectors is likely to bring benefits to Property & Casualty insurance. In manufacturing, the predictive maintenance capacity of the twins is likely to reduce overall claims in product liability and recall. Data-driven, real-time tracking of key process parameters can also help reduce product quality issues. In production facility planning (or other environments with man-machine interaction), digital twins can identify potential threats to worker, contractor and visitor safety, and even long-term ergonomics, ultimately reducing employer’s liability and worker compensation claims. In motor insurance, when applied to autonomous car features, twins can help reduce accident frequency and severity. Their application to process management or supply chain optimisation can reduce business interruption risks. Digital twins of assets can also be stress-tested against a pre-selected series of events (eg, storms, earthquakes or infectious disease outbreaks), to provide risk insights and increase resilience.⁵

Given the need for supercomputing capabilities and know-how, there is a risk that the development of digital twin software becomes concentrated among a few major players. If this were to lead to deployment of standardised software across many sectors, the risk of systemic losses in the event of software fault would rise, triggering multiple product liability claims and recalls from many manufacturers. Professional indemnity and product liability losses could be invoked for digital twin software and data providers in cases where the software or virtual assumptions are the root-cause of a fault event. 

In cases where digital twin technology is integrated with the Internet of Things (IoT) and artificial intelligence (AI), there is risk of systemic losses if AI is used in real-time to provide decisions on operation of the IoT assets. This risk is high since, due to their complexity, AI decision-making processes are often untraceable. Apart from losses due to asset failure, directors and officers claims could also be triggered for losses incurred on account of management making wrong decisions informed by the outcomes of digital-twin simulation exercises. 

Creating digital replicas also raises questions around privacy and cybersecurity. The operation of many digital twins is based on information provided by a multitude of sensors that track real-world data and movements. While they enable collaboration across processes and systems, the sensors also access sensitive information.⁶ This leaves businesses vulnerable to cyberattacks, data theft and sabotage. There are also challenges in defining data ownership and intellectual property rights, especially in the case of third-party providers of digital twin software. This is particularly relevant for insurers that rely on access to data from virtual product testing to price risk.