Engineering mechanics-aware digital twin for pipeline condition assessment

Pipeline inspection gauges are popular robots employed to inspect pipeline internal conditions. However, it has been a challenge to evaluate the mechanical performance of degraded pipelines based on inspection results. This paper presents a mechanics-aware digital twin framework that enables efficient assessment and prediction of the mechanical performance of degraded pipelines through automating the construction and updating of mechanical models of pipelines based on vision inspection data. The framework integrates machine learning-based computer vision, three-dimensional reconstruction, digital twinning, finite element analysis, damage prediction, and data-based decision-making. The framework was implemented into pipe testing, and a low-cost pipeline inspection robot was utilized to collect vision data. The results revealed that the presented approach provided accurate predictions of the stresses and failure risks of a degraded pipe. This capability supports a promising shift to inspection data-based proactive maintenance with the potential to reduce maintenance cost by over 50%.