A Method Based on SVM Algorithm for Wellbore Collision Monitoring: Using Vibration Signal Characteristics of Bit Drilling in Different Mediums

Abstract

During well drilling operations of offshore oil and gas exploration, the progress that bit drills through rocks, steel casing and other mediums produces vibration signal with different characteristics. In this work, we presented a wellbore collision monitoring method that using vibration signal produced by bit penetrating in different drilling mediums to identify those unwanted cases that bit is colliding into adjacent wellbores in drilling operation. Firstly, experiments have been carried out to acquire vibration signal of bit drilling in sorts of rock, cement and steel casing mediums. Then, to dig out signal feature in different time scales, empirical mode decomposition (EMD) method was used to decompose every signal sample into several intrinsic mode functions (IMFs) and we extracted and analyzed characteristics in time and frequency domain of all signal IMFs. Finally, support vector machine classifiers were trained with feature vectors of a part of signal IMFs to realize the function of drilling mediums identification and classification. SVM Performance test results indicates that correct identification rate of those classifiers can basically reach 90%. The method presented in this paper proved to be feasible to provide a new approach to monitor wellbore collision risks for offshore drillings.