Stuckpipe Prediction Based on Artificial Intelligence and Real-Time Sub/Surface Parameters Indicators Measurements

Abstract

Abstract Drilling operations should be as conservative as possible. These operations will become more critical in extensively depleted reservoirs. The drilling crew have to manage the allowable surface parameters in order to ensure smooth operations and can no longer rely on traditional operations either balanced or overbalanced drilling. This paper presents a model to predict stuckpipe incidents based on supervised machine-learning and real-time feature-engineering integrated with rig physical parameters. It avails of the real-time drilling data collection, fixes data issues, and analyzes the patterns within these parameters values to identify the warning signs of stuckpipe incidents. It has been tested on both types of wells either stuck or non-stuck incidents. The model was able to give alarms of downhole incidents early before the stuckpipe incidents happen to allow for time window and help the rig crew go for proactive actions to mitigate these risks.