Artificial Intelligence (AI) could be the next major technological leap that drives human technological prowess, and the EU-funded SERENA (VerSatilE plug-and-play platform enabling remote pREdictive mainteNAnce) project is at the forefront of this gathering revolution. Specifically designed for manufacturing, its platform is based upon four key technologies: remote condition monitoring and control, AI condition-based maintenance, Augmented Reality (AR)-based tools for remote assistance and human operator support, and finally, a Cloud-based platform for versatile remote diagnostics. “In summary, we have developed a distributed, lightweight and scalable Industrial Internet of Things (IIoT) platform which, through the collective use of its integrated services, will provide predictive maintenance solutions to ‘shop floor’ personnel,” states Massimo Ippolito, SERENA project coordinator.
Introducing you to the SERENA system
The platform uses a lightweight micro-services architecture, utilising Docker containers to wrap the offered services into deployable units. The SERENA AI components for predictive analytics, both distributed and centralised, are used for estimating potential failures in manufacturing equipment, allowing for the planning and scheduling of maintenance activities in a specific timeframe, thus, ensuring that the overall production process within the factory is not interrupted. On top of this, the SERENA system facilitates maintenance personnel remote support via VR/AR-based technologies that assists them with assessing the status of machinery and overall equipment within the factory. These are accessed through the use of smart glasses, smartphones and tablets.
Whilst these innovations show the SERENA platform is indeed very promising, it wasn’t always plain sailing for the project team. “It became evident from the first phases of the project that predictive analytics is a matter not only of data availability but of data of the proper quality as well,” explains Ippolito. “In case the data does not contain the features that correspond to the potential failure [of equipment], then they are of no use.” In addition, there is no single rule for every case. Human analysis and complex correlations are required to extract meaningful correlations as well as the expert’s knowledge. “Nevertheless, SERENA moved a step beyond those limitations by introducing a self-assessment mechanism and a methodology that could potentially be generic enough to capture a variety of problems,” Ippolito adds. Finally, another set of challenges encountered and conquered was the need to deploy the SERENA system in versatile environments with different connectivity capabilities and legacy systems – the difference for example between a state-of-the-art Apple iPhone-producing factory and a centuries-old factory specialising in a traditional industry. Hence, the SERENA system had to be designed and implemented to be able to support many versatile cases – and thus the system supports deployment via the Cloud or physically on the premises.
Looking to the future
Regarding the future, Ippolito strongly believes that overall, now is the time that commercially viable AI-based solutions will gradually begin to filter through onto the market. “In the future, it is expected that AI applications will keep extending into new areas providing interpretable results of increased accuracy and reduced response time,” he predicts. And finally, what of the future of the SERENA team, as the project officially ends in September 2020? Ippolito is rather coy about this but does confirm that activities to disseminate the results will continue and future initiatives to extend the SERENA system and its myriad solutions are under investigation – so watch this space.
SERENA, Artificial Intelligence, AI, Augmented Reality, AR, Cloud, manufacturing, predictive analytics, Industrial Internet of Things, IIoT, micro-services, plug-&-play