Machine learning approach for intelligent prediction of petroleum upstream stuck pipe challenge in oil and gas industry

Oil and gas are the major commodities that are fueling the advancement of human civilization. Drilling oil and gas well is a very challenging task and requires deep knowledge of the wellbore behavior to complete drilling operations successfully. A stuck pipe problem occurs when the drill pipe is unable to reciprocate and rotate inside the wellbore. A stuck pipe incident is a non-productive time (NPT) of drilling operations which is one of the major reasons for overrunning the budget of the project. This paper mainly focuses on various data-driven machine learning (ML) algorithms for the prediction of stuck pipe incidents along with the detailed discussion of various stuck pipe mechanisms and major significant drilling, well fluid, and formation parameters to enhance the performance of prediction models. According to the previous researches, traditional methods that were used to process fast high-dimensional data to predict stuck pipe incidents were inefficient, but these data-driven machine learning algorithms are capable of handling such data to extract meaningful patterns from the data. It has also been observed that the accuracy of the predictive models depends on the parameters selected for analysis. Hence, in this paper, some major parameters that are not incorporated by the previous researchers such as formation type and geomechanical stresses values are thoroughly discussed. These parameters play an important role in maintaining wellbore stability and can help predict stuck pipes using ML techniques.