A Critical Change Point Detection Method in Threaded Steel Pipe Connection Processes Using Two Stage Sequential Piecewise Linear Approach

Abstract

Leak tightness is one of the key quality characteristics of oil pipelines. In pipe connection processes, this quality characteristic is mainly characterized by two critical interpretable change points in the torque signals collected by sensors mounted on the connection machine. However, because of various noises from the operation and measurement systems, latent process factors, such as mechanical return difference, assembling misalignment, and straightness of pipes, cause various nonlinear patterns to exist in the torque signals. Hence, precisely identifying the change points for automatic quality examination is still challenging. In this paper, a two-stage modeling framework is proposed to utilize sequential change point detection to precisely locate the two critical change points. A two-phase regression model based on the F maximum test is employed to detect all potential change points in the first stage. Subsequently, a two-step backward change point selection algorithm based on mechanical principles is implemented to select the critical change points in the second stage. Finally, the change point selection based on a three-phase regression model is developed. The efficacy of the proposed framework is validated by a case study on a real threaded steel pipe connection process.

Juan Du

Assistant Professor

My research interests include knowledge-infused data science for quality improvement, industrial data analytics and machine learning, and system informatics and control for manufacturing applications.