An evolutionary decision support system for the top event early detection

The analysis of the safety of industrial processes (PHA) is an important activity carried out for design purpos-es and to satisfy regulatory requirements. Traditionally, it is based on a theoretical model of the real system with well-defined operative and environmental conditions in which system equipment, instruments and com-ponents are assumed fully reliable. In real-life situations, this is not always true, therefore new strategies must be considered to represent closely the actual state of a system. The application presented in this paper has been conceived in collaboration with a refinery plant, considering the risk scenarios associated with a fractioning column of the refinery. The main idea is to update the formal structure of the HAZOP report with the real-time data received from the plant distributed control system (DCS), with the aim to realize an evolutionary decision support system able assist the operators with the iden-tification of the most probable risk scenarios, that can happen during the evolution of the industrial process. The pilot application here presented was implemented with an electronic spreadsheet tool (Microsoft Excel©). The spreadsheet features elements that allow to extend the application for several purposes: from training pur-poses to the quantitative assessment of failure events.