Developing leading indicators-based decision support algorithms and probabilistic models using Bayesian network to predict kicks while drilling

Abstract

Abstract Predicting a kick timely and efficiently is often a challenging task due to the complexities of drilling and other well intervention activities. This work proposes a leading indicators-based approach to assess drilling operations for predicting kicks and preventing blowouts. A cause-based methodology is proposed to develop sets of leading indicators for different categories and organizational levels. Leading indicators are divided into two broad sections – real-time indicators and long-term organizational safety performance indicators. With the real-time indicators, various decision support algorithms are developed which would help to understand a kick progression scenario effectively. Probabilistic barrier failure models for different stages of drilling are also developed to assess performance of primary well control barrier – hydrostatic head. To predict barrier failure events effectively, Bayesian network models are developed combining organizational, operational and real-time indicators. The probability distribution for observing changes in real-time parameters when a kick is in progression are also determined. This study would allow both predictive (causes to effects) and diagnostic (effects to causes) reasoning of kicks and blowouts for better understanding of well control system while drilling. Developed risk models enable informed decision making with a relatively clear picture of the risk of barrier failure and provide useful information on actions required to prevent escalation of well control events.