A Multistate Bayesian Network for Accounting the Degradation of Safety Barriers in the Living Risk Assessment of Oil and Gas Plants

Abstract

The BAseline Risk assessment Tool (BART) is currently used by the Eni oil and gas company for the living risk assessment of oil and gas upstream plants. BART combines a simplified Quantitative Risk Assessment (QRA) with a Bow-Tie (BT) approach. In this work, we implemented in BART the capabilities for considering the degradation of the barriers, which affects the safety performance. For this, we resort to a multistate Bayesian Network (BN), which maps the BT of BART and whose nodes correspond to the safety barriers, each one characterized by a Health State (HS) and by a Failure Probability (FP). HS is assessed on the basis of specific Key Performance Indicators (KPIs), whereas FP is quantified from failure datasets (for technical barriers), Human Reliability Analysis (HRA) (for operational and organizational barriers) or the Analytic Hierarchy Process (AHP) based on expert elicitation (for barriers for which data are lacking). The proposed BN approach is applied to the barriers designed for limiting the consequences of a release in the slug catcher (i.e., Flash Fire (FF), Jet Fire (JF), Pool Fire (PF), Explosion (EX) or Toxic Dispersion (TX)) of the upstream onshore plant. The results of the assessment are benchmarked with those obtained with the original BART and show that the BN approach adds the capability of providing an accurate and updatable description of the barrier conditions in the risk assessment of the plant during its life.

A Novel KPI for Continuously Monitored Safety Barriers based on Probabilistic Safety Margins

“…The case study is the same as that of the companion paper (Di Maio et al, 2020a) and regards the slug catcher unit of the onshore upstream plant. The slug catcher receives the…”

“…The BN shown in Figure 1 is used to model the system and its safety barriers (see the companion paper (Di Maio et al, 2020a) for further details). The pivot node "Slug catcher LOPC", fed by the preventive safety barriers nodes, leads to the intermediate nodes describing the ignition escalation, fed by the recovery barriers, and ultimately to the consequence node.…”

“…In the BN, then, each safety barrier corresponds to a node, each one characterized by Health States (HS) and the corresponding Failure Probability (FP). Different approaches can be adopted for evaluating HSs and FPs of different types of barriers and with the different types data available (see also the accompanying papers (Di Maio et al, 2020a) and (Di Maio et al, 2020c) for further details).…”

“…These values are introduced into the BN in order to fill the CPT of the PCS, as shown in Figure 3. Figure 6 shows the complete BN for the risk assessment of the system, where the CPTs of the preventive safety barriers (including the PCS just characterized) are in place and their effect on the consequence probability is considered (see the companion paper (Di Maio et al, 2020a) for further details). Assuming that the safety barrier is monitored for the following year as well, collecting an additional time series of data (plotted in Figure 7), we are able to compute a new PSMU and a new PSML.…”

“…The performance of safety barriers has been assessed resorting to simplified risk indexes (Cozzani et al, 2009), Monte Carlo simulation (Abdolhamidzadeh et al, 2010) and Bow-Tie (BT) diagrams (Cherubin et al, 2011); in all these cases, the influence of the safety barriers current health state on their performance has been neglected (Landucci et al, 2016). In the accompanying paper (Di Maio et al, 2020a), a multistate Bayesian Network (BN) has been proposed to consider the contribution of the safety barrier health state on the risk assessment. Risk measures are periodically updated based on the Knowledge, Information and Data (KID) (Zio, 2016) that can be gathered from different sources, e.g., by monitoring system and component parameters, collecting field data, reporting inspection and maintenance information, etc.…”