Machine learning-empowered automatic analysis of distributed fiber optic sensor data for monitoring coincident corrosion and cracks in pipelines

Coincident crack and corrosion pose risks to pipelines and challenges for condition monitoring. This paper presents a machine learning-empowered approach for automatically analyzing strain data measured from distributed fiber optic sensors for monitoring coincident cracks and corrosion, which simultaneously influence distributed sensor data. This approach has been implemented to detect, locate, and discriminate coincident cracks and corrosion. The performance of the approach has been evaluated through laboratory experiments using steel pipelines equipped with distributed fiber optic sensors, considering factors such as spatial resolution and sensor deployment methods. The experimental results showed that the proposed approach achieved high mAP@0.5 (0.935) and F1 score (0.920) in detecting and locating coincident cracks and corrosion, and less than 0.009 s in analyzing a strain profile with more than 500 data. This research provides valuable insights into real-time monitoring of interacting anomalies and addresses the practical data analysis challenges associated with massive sensor data analysis.