Data learning and expert judgment in a Bayesian belief network for aiding human reliability assessment in offshore decommissioning risk assessment

Abstract

Decommissioning of offshore facilities involve changing risk profiles at different decommissioning phases. Bayesian Belief Networks (BBN) are used as part of the proposed risk assessment method to capture the multiple interactions of a decommissioning activity. The BBN is structured from the data learning of an accident database and a modification of the BBN nodes to incorporate human reliability and barrier performance modelling. The analysis covers one case study of one area of decommissioning operations by extrapolating well workover data to well plugging and abandonment. Initial analysis from well workover data, of a 5-node BBN provided insights on two different levels of severity of an accident, the ’Accident’ and ’Incident’ level, and on its respective profiles of the initiating events and the investigation-reported human causes. The initial results demonstrate that the data learnt from the database can be used to structure the BBN, give insights on how human reliability pertaining to well activities can be modelled, and that the relative frequencies from the count analysis can act as initial data input for the proposed nodes. It is also proposed that the integrated treatment of various sources of information (database and expert judgement) through a BBN model can support the risk assessment of a dynamic situation such as offshore decommissioning.

Keywords

• Bayesian belief network
• Human reliability assessment
• Expert judgement
• Data learning