Calibrating Automated Event Detection Algorithms for Real-Time Wellbore Stability Applications

Abstract

Improvement and automation of real-time drilling operations obtained significant attention during the last few years, aiming to substantially reduce non-productive time and increase safety. Wellbore instability is one of the operational challenges that can be mitigated by proactive analysis and interpretation of logging-while-drilling (LWD) and measurements-while-drilling (MWD) data. That data can be used to detect events such as kicks or losses as potential drilling complications. Successful implementation of supportive automated systems requires proper management of generated alarms to deliver reliable results. Minimization of false alarms and the classification of alarms according to their severity are necessary to provide a capable system. Following an introduction into drilling automation and wellbore stability analysis, this paper presents different approaches for the automatic analysis and interpretation of formation evaluation logs for the detection of drilling events. These algorithms were extensively tested on more than 30 data sets. The approaches are beneficial for monitoring the wellbore conditions and to alerting workers about potential overpressure regions and breakouts in the formation around the borehole. Test observations were used for developing an automated alarm and advice generation system. Alarm generation experience from within other communities and its comparison to the oil and gas industry is highlighted. Automated algorithm and alarm generation test results provide insight into the possibilities and limitations currently encountered in the application of LWD / MWD logs for the automated detection of drilling challenges. Testing has proven to be essential for identifying operating constraints that must be addressed to minimize erroneous alerts.