(55af) Enhancing Process Safety: Early Deviation Detection and Root Cause Analysis through AI for Ensuring Resilience in Complex Engineering Assets and Systems

In the context of process safety in dynamic industrial settings, the integration of new techniques for operator decision support have become paramount in recent years due to the increasing automation and digitalization of sites. A core necessity in a range of practical contexts, especially in high-risk industrial sectors such as oil & gas, chemical processing, etc., relates to the sense-making of deviation data and using it as the basis for understanding potential propagation scenarios, prediction of consequences, and prevention of unwanted events. From that perspective, this paper elaborates on the utilization of AI using semantic graphs that represent entities and their relationships, and how it can allow a deeper understanding of dynamic processes where context changes with the flow of various conditions. Moreover, the paper explores how this approach, combined with advanced data-driven solutions, has the potential to enhance overall situation awareness to a higher level so-called Advanced Situation Awareness (Ad-SA), that can provide the capability to detect deviations at an early stage and guide operations in abnormal situations.

One key aspect of this approach is the ability to track variables with the same dimensions as they propagate through the various layers of a dynamic process. This not only can facilitate the detection of deviations but also can enable the tracking of their locations and root causes connecting them to the specific components or subsystems, and predicts their propagation paths, contributing significantly to a proactive response. An appropriate level of granularity can aid in connecting early process information to improving process safety. So, in essence, the semantic graph modeling framework has the potential to enhance Contextual knowledge (CK) as a foundation for Advanced Situation Awareness (Ad-SA) which is significantly crucial in understanding the causality of deviations.

In an ever-changing environment where undesirable consequences of deviations can be disastrous, such a Hybrid approach can add value by improving early detection mechanisms, enhancing sensemaking of uncertain behavior and unknown conditions, strengthening informative decision-making, and most importantly, mitigating potential risks. The ability to identify unusual patterns and perturbations within complex industrial processes equips human operators with an enhanced level of vigilance towards ongoing context and an acceptable level of sensitivity to capture early warning signals. This Hybrid approach can act as an active guard against potential major accidents and ensures safer and more resilient industrial assets and systems.