In any new drilling project, the main objective is always to maximize return on investment. New technologies continue to be developed and used in an effort to enhance production and reservoir recovery. In this context, multilateral systems is one of the technologies that has experience significant development and application in onshore and offshore oil drilling projects in recent years. In multilateral systems, two or more horizontal wellbores are drilled from single parent wellbore, enabling drainage of multiple reservoir targets. Many are the benefits from the application of multilateral technology such as production of non-explored reserves located near existing wells and efficient reservoir drainage.

Although the obvious advantages, multilateral systems are still a technology in development stage and as such have limitations. One of the critical problems directly impacting risk and the economical viability of a project is the reliability of equipments, operations, and the human performance and interaction with multilateral systems. Petrobras, the Brazilian oil company, has been progressively adopting multilateral technology for onshore and offshore drilling projects. Aware of the criticality of reliability aspects to the risk of drilling operations in terms of possible loss of the main wellbore due to drilling problems, an intensive research work has been undertaken in that direction. In this paper, we present and discuss a methodology for probability assessment of the construction process of multilateral wells. The methodology is hybrid in nature, i.e., it combines the use of event and fault trees with Bayesian belief networks. The construction process of multilateral wells is characterized by the use of sophisticated equipments, complex operations, and human-machine interaction. As such, the construction process is initially broken down into twenty three main phases and then analyzed by means of Hierarchical Task Analysis in order to characterize the human-human and human-machine interactions. Next, event sequence diagrams are developed where the pivotal events correspond to equipment failures as well as human errors. In order to overcome typical limitations of second generation human reliability models such as unrealistic independence assumption among events, the hybrid methodology employs a Bayesian Belief Network (BBN) approach to model and quantify the probability of human error in performing multilateral related activities. The resulting BBN models not only depict the cause and effect relationship among human operators, but also the relevant performance shaping factors as well as related equipment failure for a given multilateral operation.