A Real-World Investigation of TwinSVM for the Classification of Petroleum Drilling Data

Crude oil and natural gases are extracted from underground natural hydrocarbon reservoirs through wellbore holes. Wellbores are drilled from the surface of the earth to the underground reservoirs during the drilling operations. Real field drilling operations are complex processes, which suffer from several problems such as drill pipe stuck up, wellbore instability, shale caving, sand, and loose formations etc. Drill pipe stuck up is one of the most common complications during the drilling process, which can be avoided by its early detection and identification. However, detection of drill pipe stuck up is a challenging multiclass pattern recognition problem that involves high dimensional, nonlinear, noisy, and imbalanced data. This paper proposes a Twin support vector machine (TwinSVM) based learning model for automatic detection and identification of stuck pipe events. The proposed model has been implemented on the MATLAB platform and tested on five different datasets acquired from an Iranian oil field. The hyperplane parameters of TwinSVM are optimized using Genetic algorithm (GA). The performance of the TwinSVM model is compared with six additional classifiers including artificial neural networks and support vector machine. The test outcomes, analyzed and presented in the paper, clearly indicate the superiority of TwinSVM over other classifiers used for identification of pipe stuck up events.