Research on Evaluation Method of Make-Up Torque of Premium Thread Based on Logistic Regression

Abstract

AbstractSustained casing pressure (SCP) caused by premium thread leakage is mostly considered as the common failure form of well integrity in gas wells, where the make-up torque more likely poses a serious threat to sealing performance. At present, most of the evaluation methods for the sealing performance of premium thread are based on finite element analysis, laboratory test which mainly focuses on the stress distribution of torque shoulder under different load states. However, the influence of on-site operation quality is not considered. This paper presents a machine learning method based on logistic regression for assessing the sealing performance of premium thread in gas wells. The approach includes two models, namely statistical model and line-based model. These models consider the characteristic parameters of torque data as well as data points on curves, which are used to automatically classify the make-up torque curves. Moreover, a comparison of two models performance is presented and analyzed. The line-based model exhibits better classification accuracy and response speed. Meanwhile, in contrast to previous works, these models take the field operation quality into account, which is capable of overcoming limitations of finite element analysis and laboratory tests. Afterwards, a case study based on machine learning algorithm and logging data is presented to illustrate the feasibility of the proposed approach, and also demonstrate that the assessment of sealing performance based on machine learning contributes to the quantitative evaluation of operation quality, as well as an inspection of the quality and integrity of premium thread.KeywordsPremium threadSustained casing pressureSealing performanceMachine learningLogistic regression