Knowledge graphs to convert large Safety Control Structures of modern industrial establishments

The industrial and societal landscape is rapidly changing due to technological advances and rise of automation, as well as globalization at different scales. Modern industrial establishments are complex systems, where the interactions and synergies among organizational, human, and technical elements are easily recognized (Pasman, 2009). They are labelled as Socio-Technical Systems (STSs), being characterized by a high number of elements tightly interacting, which make them complex and prone to unexpected variability and highly interdependent behaviours (Dekker et al., 2011). In such a context, the modern industrial systems demand for a paramount concern on safety. Such concern is particularly evident for all the establishments involved in the storing, handling, production, and use of dangerous substances (the so-called Seveso establishments). These latter have the potential to lead to major accidents with severe consequences on equipment and moreover on the population and the plant’s surrounding environment. The European Union ruled on the management of these plants through the European Directive 2012/18/EU (i.e., latest version of the European Directive 82/501/EEC). Accordingly, operators of Seveso establishments must adopt effective strategies to prevent major accidents and mitigate their consequences for human health, economic, and environmental damages. The management and control of Seveso establishments is a complicated and intricate process, in which several agents exchange a large amount of information, and their decisions influence the whole industrial eco-system. For this purpose, the Systems Theoretic Accident Model and Processes (STAMP) model represents a suitable alternative to support the management of such establishments considering its foundation in control theory and models in the hierarchical Safety Control Structure (SCS), where the system agents are identified and interactions between them becomes evident. While the value of STAMP and its nested techniques has been largely proven in literature (Patriarca et al., 2022), a limitation arises when modelling large complex STSs since the resulting SCSs become too complicated to gain useful insights (Nakhal A. et al., 2023). This research presents a possible solution via the adopting of Knowledge Graphs (KGs), meant to be an extension to the SCS development.