Data-driven detection and identification of undesirable events in subsea oil wells

Abstract

Condition-based monitoring (CBM) systems have gained huge popularity in recent years with technological leaps that have arisen. Sensor-technology, communication systems, and computational capability have introduced innovative systems to monitor, analyze, and identify failures in industrial plants, production lines, machinery, and equipment. The gas and oil industry lose billions of dollars yearly related to abnormal events. Thus, abnormal event management (AEM) has become their number one priority, which aims to timely detect and diagnose abnormal events so that preventive actions can be taken. Similar to AEM, this research deals with the detection and classification of faulty events in offshore oil wells by creating a CBM system. The events used in this work are a part of the 3W database developed by Petrobras, considered the world’s thirdlargest oil producer. Seven events categorized as faulty events are considered, as well as instances considered as normal operation. This work conducts three experiments. The first experiment is related to a new feature extraction strategy, while the last two experiments are related to two different classification scenarios. The proposed systems achieve an overall accuracy of 90%, indicating that the system is not only able to detect faulty events but also successfully anticipating incoming failures.