Digital twin based predictive diagnosis approach for submarine suspended pipelines

Abstract

Subsea pipelines are vital to offshore oil and gas production systems, with submarine suspended pipelines being the most prone to accidents. Therefore, predictive diagnosis of these pipelines is critical. Traditional inspection methods, though widely studied, fail to offer long-term and multi-directional diagnosis. To address this, the paper introduces digital twins for enhanced predictive diagnosis. A unit-level digital twin framework for submarine suspended pipelines is proposed, leveraging digital twins' capabilities in real-time monitoring, failure prediction, and optimization. Predictive diagnosis, however, requires insights into the pipeline's future condition. Thus, a PFD-TCN (Polynomial Fitting Denoising Temporal Convolutional Network) model is introduced, combining TCN with a polynomial fitting function for noise reduction. This model reduces prediction error by over 50 % compared to traditional TCN models, significantly improving the accuracy of pipeline strain forecasting. Importing real-time and predictive data into the digital twin enables real-time tracking of pipeline dynamics and fault evolution, facilitating further diagnostic analysis within the twin framework. The integration of fault diagnosis and digital twins establishes a novel predictive diagnosis approach for subsea pipelines. In conclusion, the predictive diagnosis system can provide a guarantee for the safe production of offshore oil and gas.