Securing Digital Manufacturing Platforms with STAR AI Technology
Manufacturing could be considered at the historical level the “Digital Twin” of human skills evolution. It reflects our growth and ability to build tools, goods, to overcome challenges. Nevertheless, it is one that very special area of our societies where human mind turned skills in complex tools. Industrial revolution and its disruptive impact over the ages is peered by other probably less observed major changes in societies. Those ones, like accessibility of education, communication and travel means, accelerate, and deliver inputs for new waves of growth, efficiency, and diversity in manufacturing. All those cycles of complexity growth refined the understanding, preventing, monitoring and reacting to threats related to safety, security and privacy of humans, machines and processes.
Communication networks and computers are key enablers of last few decades able to bring in the game a big number of process and motion control systems. If until no more than 30 years ago, subjects like Artificial Intelligence, Machine Learning or Simulation have been perceived as a subject of a rather reduced academic community we have can see now how various industries are building models of action almost completely digitalized. Control and operational data flowing through digital platforms associated to manufacturing processes are no longer managed in closed systems with no connectivity to outside world. Mixed fog-edge -cloud deployments constitute the new reality where functionalities are managed in a DevOps manner. Trust and security are now considered from some different perspectives simultaneously, each of them being relevant for both horizontal functions but also on the vertical applications views.
Up to recent time the level of explainability of decision with many industrial applications stayed within the level of “Human in the loop” control doing enablement of various processing steps (e.g. possible sensitive chemical plant processing or visual inspection of machine tools space), or being based on purely declarative rule based systems as classic Siemens Simatic controllers. Those approaches have been preferred offering a wide understanding and validation of processes to be executed. Still, analytics technologies offered a new push to Machine Learning and Deep Learning techniques peering them with descriptive technologies, like the ones related to Knowledge Graphs or Constraints Programming, both being able to instrument together problem design hypothesis and operational data collected. The aim of such techniques is to offer the tools usable for validation of execution scenarios before effective deployment in production systems.
STAR project consider not only the instrumental aspects of trustworthiness and explainability of AI but also the ways how the AI solutions do the step from the work of art to systematic industrial use. This means a specific focus on the needs to offer certification mechanism altogether with pre-validated compatibility and reliability application scenarios.
Last but not least, STAR is taking the effort to provide a solid path towards industrialized AI being aligned and contributing to normative efforts at both EU and global level.
By Cosmin-Septimiu Nechifor, SIEMENS