SRA-E Lisbon 2017

Evaluation of potential accident scenarios is of paramount importance for safety-critical industrial sectors such as chemical manufacturing. It supports design of appropriate safety measures to both prevent and mitigate their occurrence – denominated safety barriers. Specific regulations (e.g. Seveso directives) require safety reports to be provided to competent authorities before plants are built and operated. Operational support may also represent another purpose of such analysis, which benefit from the growing trend of collecting performance indicators. Progressively larger amounts of data on the performance of safety barriers are becoming available in industry on a (quasi-)real-time basis. Methods developed for “static” risk assessment, mainly aimed at supporting design, do not have the capability to process such information flows and eventually provide reliable results. However, specific risk assessment techniques may represent a suitable starting point for such perspective shift. The bow-tie analysis may be representative of the latter group. Its goal is twofold: i) identification of major accident hazards on the basis of the equipment and substances handled; ii) deep study of accident causes, probability levels and safety measures. Advanced methodologies have been developed to upgrade such methods to dynamic application based on continuous update and improvement of inputs. This contribution aims at providing an overview of the challenges faced by the current techniques and a selection of potential innovative solutions: from the dynamic update of hazard identification, to the continuous assessment of safety barrier performance based on technical, operational and organizational indicators. Furthermore, the contribution paves the way for future risk assessment techniques exploiting the recent re-birth of artificial intelligence, which would potentially allow for definitive cognition of lessons from past accidents.