Why digital twins need AI-driven testing

Digital twins continue to gather steam, with Gartner predicting the use of digital twins to triple by 2022. A Digital twin is a digital representation of a physical object or system. They are being widely used within supply chain management to track items as they move across companies, also using blockchain technology, and within automated factories to track wear and tear and perform predictive maintenance to reduce downtime.

Digital twins are also becoming synonymous with IoT as people are increasingly modeling “things” so that analysis and operations can be done with minimal calls to the outside world (e.g., reducing power consumption). For example, creating a digital twin of a thing like a freight truck allows you to have all your different fleet management, supply chain visibility and vehicle maintenance operating on the digital twin rather than all having to talk constantly to the physical vehicle with different requirements.

This digital twin can test and monitor the performance of the IoT ecosystem. However, its unique characteristics need to be considered to ensure your testing strategy is effective. Digital twins are relatively uncontrolled systems, unlike a typical software system where inputs and outputs are well established, so simply defining what is and isn’t a failure can be difficult. Also, a digital twin mirrors the physical world and, as a result, there’s a lot more variability. From temperature sensors to tire pressure sensors, there’s a vast magnitude of possible combinations of possible inputs, coupled with the business logic, and lots of independent, different parts to the digital twin.

The complexity of the digital twin means that traditional testing strategies that rely on hundreds of test cases simply will not suffice. You need to move beyond traditional script-based testing to intelligent testing driven by a combination of AI and machine learning, with auto-generated tests and learning algorithms to determine the pass or fail.

Another consideration is that digital twins are not static and move between software systems. Going back to the freight truck management example, the digital twin of a motor part may move across several companies as ownership of the motor changes. This means you need to test the compatibility of all those different software systems interacting with the digital twin. The complexity of a digital twin requires not only AI and machine learning to cover the vast number of complex scenarios, but a test automation tool that is able to test the entire ecosystem.

Digital twins are here to stay, and have the potential to transform the way products are designed, manufactured and maintained. As digital twins become more commonplace, the adoption of intelligent AI-driven testing will accelerate to understand the entire IoT ecosystem and maximize outcomes.

All IoT Agenda network contributors are responsible for the content and accuracy of their posts. Opinions are of the writers and do not necessarily convey the thoughts of IoT Agenda.