Probability of Wellbore Failure and its Prediction Using Machine Learning

Abstract

Wellbore instability (WI) is one of the major challenges experienced during drilling operations costing the oil and gas industry over $1 billion yearly. During the drilling, borehole breakout and drilling induced fractures are the two main instability problems which may lead to stuck pipe, sidetracking and loss of circulation. Due to large uncertainties of subsurface data such as rock Poisson ratio and permeability and impact of drilling fluid dynamics and density, the prediction of wellbore failure probability poses a challenge. Bayesian analyses provide a great framework for this type of probabilistic analysis. In this research, two common failure criterion-Mogi-Coulomb and Mohr-Coulomb failure criteria were applied to synthetic data. Subsequently, Gaussian Naïve Bayes (GNB) algorithm was applied to the outcome to probabilistically predict wellbore failure, and it is shown that GNB was able to predict wellbore failure with 86.7% and 67.3% accuracy respectively for the Mogi-Coulomb and Mohr-Coulomb model