Downhole incidents such as kick, lost circulation, pack-off, and hole cleaning issues are important contributors to downtime in drilling. In managed pressure drilling (MPD), operations margins are typically narrow, implying a risk of more frequent incidents and more severe consequences than during conventional drilling. Detection and handling of symptoms of downhole drilling contingencies at an early stage are therefore crucial for the reliability and safety of MPD operations.
This paper describes a method for early detection and localization of incidents, based on a fit-for-purpose model of the downhole pressure hydraulics, distributed pressure measurements from wired pipe, and statistical change detection methods. Using statistical change detection methods, it is possible to detect the incidents and their location at a very early stage when small changes in process data are covered in measurement noise, which may be very hard to observe for a human operator.
The method has successfully been tested on experimental data from a medium-scale horizontal flow loop in Stavanger, Norway. The flow loop represents a 700 m borehole with emulation of the following downhole contingencies: drillstring washout, drill bit nozzle plugging, gas influx and fluid loss. In the tests, these incidents were successfully detected at an early stage. The developed detection and localization method could be included as a diagnosis tool in a drilling system with MPD or conventional drilling.