A reference architecture to implement Self-X capability in an industrial software architecture

The recent advancements in the industrial communities which have taken place from 2011 and are nowadays referenced as “Industry 4.0” allowed the scientific community to highlight the benefits of data-based approaches when dealing with the industrial operations. The most impressive outcomes, in particular, leveraged on the tools implementing Machine Learning algorithms, which gave companies new tool to understand current and future statuses of operations and processes through a data-driven approach. These tools often rely on historical data and can indeed cluster or label events basing on existing or past measurement, while, when dealing with unforeseeable events, are limited to alert warning. In order to overcome these burdens, the concept of Autonomous Computing is nowadays gaining more and more importance, as a Computer Science discipline dealing with the resilience of a system with respect to unexpected and unregistered behaviours. The adoption of this kind of technology would hence allow to enhance the existing data analytics industrial infrastructures reaction capabilities, able to maximise the assets’ availability beyond the structural limits of traditional Machine Learning approaches. This work proposes a reference architecture to embody these functionalities in an industrial software architecture based on Artificial Intelligence processes and pipelines, to pave the way for a full implementation of autonomous responsiveness from the software modules constituting the architecture. Despite the rising interest in Autonomous Computing, its adoption in industry is indeed slowed by the fear of complexity in its implementation. The (numerically) poor body of knowledge about software architectures devoted to Autonomous Computing is hence enriched, and a reference implementation is presented as well, in order to encourage the practitioners’ community towards the adoption of this technology.