Objective  The pipe bending strain identification technology based on an inertial mapping unit (IMU) in-line detector has seen extensive applications in China and abroad. IMUs obtain bending strains along entire pipelines through a solving process, using pitch angles and heading angles acquired by a gyroscope. However, deviations from the detection path may occur during the practical in-line detector operation, due to the sizes of IMUs and other factors, leading to inevitable errors between the bending strains identified by IMUs and the actual pipeline conditions.

Methods A simulation model was developed for IMU in-line detectors based on their actual size, without abnormal vibrations and other interferences. In addition, a simulation database was created for pipelines with a diameter of 508 mm under varying bending strain conditions. Moreover, a solving algorithm optimized through an ANNExtraTree deep learning model was introduced. These tools were leveraged for error analysis and to delve into the bending strain solving algorithm. Furthermore, full-scale pulling experiments were conducted on pipeline bending strains, to verify the feasibility and accuracy of the optimized solving algorithm.

Results  This study revealed that as pipeline bending strain increased, the errors between IMU detections and true bending strains grew. The mean square error, error covariance and error standard deviation of the bending strain data derived from the optimized solving process decreased from 0.007 0, 0.004 9, 0.070 2 to 0.001 2, 0.001 6, 0.012 6 respectively, while the coefficient of determination, indicating correlation, rose from 0.443 to 0.981. Furthermore, the full-scale pulling experiments conducted to validate the algorithm revealed substantial reductions in errors between bending strains computed by the optimized solving algorithm and those detected by a strain gauge at 79.6%, 79.4%, and 76.0%, respectively.

Conclusion The proposed optimized solving algorithm is verified feasible and accurate through finite element simulations and full-scale pulling experiments. The study findings provide technical support and guidance for accurately identifying bending strains along pipelines based on IMU.