Comparative analysis of the buried trunk pipeline sections based on the data obtained at different in-line inspections showed that there is a significant group of repaired sections with non-standard values of bending radius. The purpose of this work is automated detection of pipe sections with decreasing values of bending radius despite the undertaken compensatory activities. Preliminary analysis of data of five in-line inspections showed the presence of more than 200 thousand sections, 20% of which have the non-standard bending radius. The proposed method includes selecting non-standard bending radius dataset, restoring missing measurements, clustering dataset, and assigning compensatory activities. As a result, for each thirty-kilometer part of the oil pipeline, the group of sections with the stable change in spatial position is allocated. For this group it is required to perform compensatory activities to bring the bending radius to design values with preliminary studying geological structure of containing soil.