Severe bit damage is a known issue in west Texas land drilling due to abrasive sand formation and interbedded hard stringers. Operational performance and rig cost are often impacted by penalty fee of bit DBR (damage beyond repairable), low ROP (rate of penetration) with worn bit, and inefficient decision-making on tripping. A real-time data analytics application is developed aiming to provide actionable information to operation to expedite decision making process.
A historical dataset of surface mechanics data and bit records is collected from 40 bit runs drilled in 2016 and early 2017. A hybrid data analytics procedure consisting of conventional physical modeling of drilling mechanics and supervised learning using machine learning technique is conducted to separate bit failure pattern from normal formation transition and drilling parameters adjustment. A metric based algorithm is constructed for real-time monitoring of bit drilling performance and early warning on bit cutter wear conditions.
A web-based real-time software is developed and field trialed on three wells with satisfactory results. Subsequent deployments in DART (Drilling Automation Remote Technology) center and field offices have been quickly rolled out for five rigs in west Texas. Positive feedback is generated from operation and engineers. Attributed to the success of agile development framework and adaptive software architecture, other advisory mode features such as motor life monitoring, smart-tripping evaluation, and sliding diagnosis etc. are under development.
The application discussed in this paper combines expert's domain knowledge with machine learning techniques and provides actionable information to support on-site operational decisions. The development and deployment of this application follows an agile mode innovation framework, through which operational need and technical solution are quickly bridged and tangible business value is able to be delivered in short term.