PSM Evergreening with Artificial Intelligence

Abstract Several "ad-hoc" safety assessments are routinely performed in Major Hazard Facilities such as Operational Risk Assessments (ORA), Management of Change (MOC) HAZOPs, Pre Startup Safety Reviews (PSSR), maintainability reviews and Root Cause Analyses (RCA). Recurrent safety assessments include HAZOP revalidations (re-HAZOPs), Safety Case / Safety Reports, BowTie reviews, etc. Undertaking these assessment in a coherent and systematic manner is problematic, not to mention keeping all the safety related information updated and in the same place. This paper describes how Artificial Intelligence (AI) can be used to build and maintain a "live" functional twin of any industrial facility where Process Safety (PS) information is continuously updated and accessible to support standard risk management processes, which is the basic feature underpinning the concept of "Process Safety Management (PSM) evergreening". Employing Multilevel Flow Modelling (MFM), a functional twin of the facility is created by the HAZOP Assistant software application. With its reasoning capability, the model creates a map of all possible causal-effect relationships, linking causes with end consequences and safeguards alike a HAZOP study or a bowtie. This "live" map of causal-effect relationships can be accessed and displayed at any time in the form of an event tree or bowtie. The functional twin is not a database of words but an actual physical model which includes qualitative mass and energy balances. As PS knowledge improves, the model can be easily updated with the new information. When changes are made to the facility, the model can be updated to reflect the "as-is" status, so that safety assessments are updated and relevant. The paper highlights how the HAZOP Assistant software application builds and maintains a functional twin of the facility to support the above processes consistently and efficiently, in one platform and reducing errors and costs.