Testing Machine Learning Algorithms for Drilling Incidents Detection on a Pilot Small-Scale Drilling Rig

Abstract

In recent years, drilling digitalization and automation have advanced from being automation of rig floor equipment to an idea that is starting to be applied to entire drilling processes. However it is very costly in terms of field testing and validating developed novel technologies. To address this limitation, we take advantage of a laboratory drilling rig to run a large number of drilling tests. By introducing various drilling scenarios while drilling different formations using various combinations of the operational parameters, we could be able to collect a large amount of data for data-driven methods development and testing. The main study in this article is to develop machine learning algorithms for identifying abnormal drilling and test these algorithms on the rig based on the responses of the rig sensors in real-time operations. The idea also helps us determine what the most important parameters or their combinations for drilling incidents detection are, which we could pay greatest attention to make right decisions with the help of drilling data during real-time operations.