Fluid losses are still today one of the most challenging problems in well construction. The scenarios faced by operators during development and exploratory campaigns in the deep water pre-salt area are characterized by natural fractures, vugs and caves. Therefore, problems related to loss of circulation are critical, increasing the non-productive time and consequently, well construction costs. Additionally, in several situations, conventional drilling limitations prevent the reaching of the well target. The present study proposes the application of a methodology to define optimum loss control material (cross-linked pills, fluid loss squeeze, bridging agents, cement), among those available for each situation, to minimize lost circulation events during drilling operations. An Artificial Intelligence strategy based on Supervised Learning was defined to generalize data collected from five hundred lost circulation events over a three years period. Human Computer Interaction principles were used on the development of an interface where the field engineer can interact with training data while having little to no Machine Learning knowledge. The use of empirical analysis and learning strategies as tools to assist the decision making process in the form of lost circulation countermeasures is described by this paper. The method was validated on data collected from several different wells in the Santos Basin, Brazil, pre-salt area. The strategy was already applied in two real cases resulting in a six days well construction time saving.

No abstract