Application of one point clustering algorithms to develop a defect comparison model for differential time inspection of chemical pipelines

Abstract

This study addresses the challenges of defect comparison in differential time inspection of industrial long-distance pipelines by proposing a clustering algorithm-based defect comparison model. The research focuses on the safety management needs of underground pipelines in the petrochemical industry, particularly on accurately matching defect distribution and features after multiple in-line inspections (ILI). A data-driven automated comparison procedure employing clustering analysis is developed, effectively identifying defect similarities and distribution patterns, thereby significantly improving comparison efficiency and accuracy. Notably, this work pioneers the integration of data preprocessing techniques, such as mileage calibration and weld seam alignment, with weighted feature clustering to enhance both the reliability and sensitivity of defect matching. Validation through Unity Plot analysis confirmed that the proposed method reduced manual matching errors to zero and improved comparison efficiency by 76 %. The findings demonstrate that the model not only enhances the reliability of defect matching but also provides robust technical support for the safe operation and maintenance strategies of underground pipelines, with potential extensions to multi-source data integration and predictive maintenance applications.