Statistics Pattern Cointegration Analysis-Based Bit Bounce Detection for Drilling Process

Fault detection is essential for the safety of drilling process. Without proper detection and appropriate treatment of bit bounce, it may result in serious drilling accidents. The data distribution shift will lead to the nonstationary property. Yet the statistics of drilling parameters may have similar equilibrium relationships across different depths due to the existence of automatic drilling control system. To solve the issue of distribution shift, a statistics pattern cointegration analysis (SPCA)-based bit bounce detection method for drilling process is proposed. The statistics pattern of drilling parameters is constructed. The cointegration analysis (CA) is conducted to discover equilibrium relationships among nonstationary statistics for process monitoring. The corresponding alarm deadband is designed through the analysis of alarm duration and alarm deviation. An industrial case study from an actual drilling process is conducted to demonstrate the effectiveness of the proposed fault detection method. The proposed statistics pattern cointegration analysis approach is conducive to maintaining a satisfactory fault detection performance in the case of distribution shift in the industrial process.