APPLICATION OF MACHINE LEARNING METHODS TO PREDICTION OF HOLE CLEANING FROM CUTTINGS

Abstract

Drilling operations typically face several issues that can significantly impact a drilling program by increasing the time required to drill to a given depth. Such problems include pipe sticking, and the elimination of such accidents is the most difficult, the most time-consuming, the most risky, and even the entire wellbore or part of it can be abandoned. This problem is closely related to hole cleaning conditions and can be overcome by providing good hole cleaning conditions. Wellbore cleaning problems can be reduced by using machine learning models that can predict wellbore condition with real-time field data and suggest the best course of action. By leveraging the power of machine learning, well drilling companies can achieve more accurate and efficient hole cleaning estimation, resulting in improved drilling productivity. In this work, two machine learning models were trained: the first model is designed to predict the presence and occurrence of a cutting bed in a well during drilling; the second model is a regression model that can be used to predict the cutting bed height and helps to show how serious the situation is at the bottom of the well or if it is not critical. The results of this work can be used by drilling engineers to receive real-time notifications about the risk of insufficient well cleaning during drilling.