Reliability assessment of offshore oil multilateral wells in Brazil

This paper provides a procedure for the reliability assessment of multilateral wells construction process. The modeling approach is hybrid in nature as event sequence diagrams are used to model the sequence of events, while the pivotal events are characterized in terms of Bayesian belief networks (BBN), for cases where polichotomy and dependent variables are required, and by fault trees otherwise. In order to quantify the system failure probability, the resulting hybrid model is converted to its equivalent BBN representation and evaluated via a junction tree algorithm. The procedure is demonstrated by a numerical example in the context of an ongoing shallow water offshore project in Brazil, where, in particular, the pivotal events related to human actions are modeled via Bayesian belief networks