Safety analysis of offshore decommissioning operation through Bayesian network

Research output: Contribution to journalArticle

1 Citation (Scopus)
9 Downloads (Pure)

Abstract

Decommissioning of offshore platforms is becoming increasingly popular. The removal of these heavy steel structures is characterised by high risks that may compromise personnel safety and loss of assets. The removal operation relies on dedicated barges and heavy lift vessels that may descent or capsize because of mechanical or structural failure. The knowledge of associated hazards is driven by experience and failure data are often obtained empirically through analogous operations, which further introduces uncertainty to the risk analysis. This paper proposes an integrated safety analysis approach for conducting a decommissioning risk analysis of offshore installations. The approach incorporates hierarchical Bayesian analysis (HBA) with Bayesian network (BN) to assess the accident causations leading to futile decommissioning operation. First, the overall system failure of a lifting vessel was reviewed with an emphasis on where safety issues arise. In addition, the failure data obtained from expert judgements were aggregated through statistical distribution based on HBA. The aggregated failure data are then used to conduct dynamic safety analysis using BN, to assess and evaluate the risks of offshore jacket removal operations. The accident model is illustrated with a case study from Brent Alpha decommissioning technical document to demonstrate the capability of incorporating HBA with BN to conduct a risk analysis.

Original languageEnglish
Pages (from-to)99-109
Number of pages11
JournalShips and Offshore Structures
Volume15
Issue number1
Early online date21 Mar 2019
DOIs
Publication statusPublished - 2 Jan 2020

Keywords

  • Bayesian networks
  • decommissioning
  • safety analysis
  • offshore jacket structures
  • hierarchical Bayesian analysis

Fingerprint Dive into the research topics of 'Safety analysis of offshore decommissioning operation through Bayesian network'. Together they form a unique fingerprint.

  • Cite this