Early Stuck Pipe Detection Based on Time Series Analysis

Abstract Recognizing stuck pipe signs in an earlier time and adopting appropriate action can greatly reduce non-productive time during drilling and improve drilling safety. The purpose of this paper is to propose a time series analysis approach and build a reliable and easy-to-use tool to automatically detect stuck pipe accurately and early. Based on the in-depth theoretical analysis and historical stuck pipe data analysis, main early stuck pipe indicators during different drill operations are identified. More than ten time series analysis algorithms and machine learning algorithms are utilized to capture subtle change trends and local change patterns in noisy real-time measured surface data during drilling, and a self-adaptive threshold value determination method is proposed to improve the detection accuracy under different data measurement quality. Moreover, correlation analyses are conducted on the multidimensional time series based on the studied rules and priori knowledge to reduce the false alarms caused by curtain drilling operations. An early stuck pipe detection software tool is built. The tool consumes both real-time and archived drilling data, and provides alarms when stuck pipe indicators are detected. 19 wells’ historical drill data are fed to the tool for algorithm verification, the results show that the tool can automatically detect all 22 stuck pipe incidents in the data set, among which, 5 incidents were detected 2 hours before the field recorded event time, 2 incidents were detected more than 1 hour before the field recorded time, 8 incidents were detected 3-42 minutes before, 4 incidents were detected about the same time as the field recorded time, and 3 incidents were detected less than 2 minutes later than the field recorded time. The average false alarm rate is about once per 48 hours operation time. The tool was also connected to the real-time drilling data in remote drilling operation support center. During the trail, two early stuck pipe signs are captured accurately. The computation time for each real-time data point is less than 0.5 second, which meets the real-time requirement. According to the demonstrated stuck pipe detection rate and false alarm rate, the tool is promising and beneficial for detecting stuck pipe in early time automatically, which improves the drilling safety and reduce non-productive time.