Objective Elbows, as geometrically discontinuous structures in buried oil and gas pipelines, are susceptible to failures such as fatigue damage, local buckling, and stress-corrosion cracking due to stress concentration. Accurate identification of their location and geometric features is crucial for pipeline integrity assessment and safe operation. However, current in-line inspection technologies face significant challenges in reliably identifying elbows, determining their types, and characterizing their spatial geometry. Therefore, developing a high-precision elbow identification method based on multi-source data fusion is imperative.

Methods A method for elbow identification and spatial feature analysis was developed based on data from the Inertial Measurement Unit (IMU) and odometer. Continuous pipeline attitude information was obtained from IMU and odometer data collected by the in-line inspection tool, then filtered and resampled using mileage constraints. Changes in attitude angles between consecutive sampling points were combined with pipeline segment geometric parameters to eliminate interference from girth welds, enabling identification of hot-bent and cold-bent elbows. Finally, the local three-dimensional spatial trajectory of the elbow region was reconstructed from attitude angle and mileage data to characterize its geometric features.

Results The proposed method was applied to a 4 400 m small-diameter gathering pipeline. Comparison with traditional magnetic flux leakage (MFL) inspection revealed that the new method identified 151 elbows (117 hot-bent and 34 cold-bent), whereas the traditional MFL inspection identified 137 elbows. Of these, 128 elbows were matched between the two methods, achieving a 93.43％ coincidence rate based on MFL inspection results. While maintaining high consistency, the new method also provided elbow type identification, which the traditional MFL inspection could not achieve. Additionally, the three-dimensional spatial trajectories of 23 newly identified elbows were reconstructed, showing continuous bending characteristics consistent with typical elbow geometry. In contrast, all elbows identified solely by traditional MFL inspection were located in girth weld areas and lacked continuous bending characteristics, demonstrating that three-dimensional trajectory analysis effectively eliminated girth weld interference and accurately determined elbow types.

Conclusion The IMU-based elbow identification method offers clear advantages in detection accuracy and geometric characterization. It supports defect localization, quantitative evaluation of long-distance pipelines, trenchless inspections, and the development of operation, maintenance, and repair plans.