Research on Automatic Classification of Premium Threaded Connections Make-Up Torque Curve Based on CNNs with Data Augmentation

Leakage of premium thread connections tubing is the main reason for annulus pressure and affecting well integrity level. At present, helium gas seal detection and manual monitoring of make-up torque curve are mainly used to ensure the integrity of gas seal of well string. However, the helium seal detection environment is static detection, which fails to describe the air tightness of the tubing under complex downhole load; The manual monitoring of the make-up torque curve depends on the field engineer with certain experience to check the standard curve one by one. The results are greatly affected by subjective factors, so it is difficult to unify the measurement standard. Therefore, a machine learning method based on convolutional neural network (CNN) is proposed to automatically identify and classify the makeup torque curve of special threaded tubing. In order to achieve this goal, firstly, according to the failure of gas seal detection, manufacturer's manual and field experience, the categories of makeup torque curve are divided, including typical curve, acceptable curve and unacceptable curve. Secondly, in order to improve the model training accuracy and further improve the prediction results, the data expansion technology is used to expand the training database. Finally, the multilayer convolutional neural network model is built and trained and verified based on the data. In the model verification stage, four comprehensive evaluation methods are used: typical rate (TR), acceptable rate (AR), unacceptable rate (UR) and accuracy (P). The proposed CNNs model is evaluated accurately and compared with state-of-art machine learning algorithms such as SVM and logistic regression.