Inertial Measurement Unit (IMU) bending-strain data play a critical role in detecting and assessing landslide impacts on pipelines. Inline inspection (ILI) vendors often flag many pipeline bending-strain features based on IMU data; however, most identified features (>90–95%) are not associated with geohazards such as landslides. Manually reviewing these strain indications is slow, labour-intensive and relies heavily on the experience and expertise of the analyst. This study describes the development and testing of a machine learning (ML) method to automatically screen and prioritize IMU bending-strain anomalies linked to landslide-related deformation.

A Convolutional Neural Network (CNN) deep-learning model was trained using IMU data from North American transmission pipelines. Input features included horizontal and vertical bending strains, horizontal out-of-straight, pipe elevation, and topographic elevation. Model performance was evaluated in collaboration with a major North American pipeline operator, using a hold-out test dataset. The model achieved approximately 88% accuracy, 80% recall, and 90% specificity – in other words, it filtered out approximately 90% of benign features while capturing over 80% of landslide-related deformations.

This research and pilot application demonstrates the potential of ML models applied to IMU data. The model already has the potential to reduce analysis time per site and improve consistency in identifying high-risk features. By rapidly identifying landslide-related pipeline deformation, it helps operators allocate resources more effectively, respond faster to emerging threats, and maximize the value gained from their IMU data. The future integration of additional contextual data such as lidar and landslide mapping, along with increases in the training dataset size, are expected to yield substantial future improvements to model performance.