Hybrid dynamic risk modelling for safety critical equipment on offshore installations

Abstract

Safety-critical equipment (SCE) are important safety barriers installed on offshore installations to prevent the occurrence as well as mitigate the consequence of major accidents. Risk assessment is significant for SCE in consideration that the risk of major accidents is increased due to the harsh operational environment and the hazards introduced by periodical preventive maintenance (PM). However, it is difficult to access risk precisely or robustly as the dynamic characteristics of SCE degradation and the influence of human error can increase the uncertainties. In view of this, this study proposes a new hybrid dynamic risk modelling methodology for SCE on offshore installations to capture the causality and dynamic dependencies precisely, where dynamic Bayesian network (DBN) technique and support vector regression (SVR) algorithm are combined. The impact of human error is evaluated through a complex structure modelling of risk influence factors (RIFs) in DBN. The hybrid SVR-DBN methodology is applied in the case study to assess the dynamic risk profile of a typical SCE (pressure safety valve, PSV) in the context of operation and maintenance respectively. The effectiveness of the hybrid SVR-DBN methodology is verified and the synergy from the integration of SVR and DBN overcomes the limitations of traditional DBN for modelling actual dynamic dependency between dynamic nodes in jacent time-slices. Besides, the modelling of RIFs introduced in DBN facilitates the assessment of human error. The hybrid SVR-DBN methodology is a systematic tool for the dynamic risk assessment of SCE on offshore installations in consideration of both the dynamic characteristics of SCE degradation and the influence of human error.