Dynamic Stress Analysis of Pipelines through Finite Element Modeling and Machine Learning-based Electrical Impedance Tomography

Abstract

Pipelines are one of the most reliable means for transporting oil and gas from production to consumption sites. Pipelines are at the risk of degradation over time due to different types of defects, boundary conditions, and loading conditions. To ensure the safe and reliable transportation of oil and gas products through pipelines, in-line inspection (ILI) is widely used for the condition monitoring of pipelines. The in-line inspection process relies on localized measurements within the pipeline at a very high speed and it can impose substantial forces and vibrations on pipelines, particularly, the ones with integrity-concerning anomalies. Furthermore, depending on the geographical conditions, environmental considerations and maintenance requirements, pipelines often pass through excavated sites, bridges, water, and muskeg/bogs. These partially supported, excavated, floating and above-ground pipelines exhibit excessive vibrations compared to their buried counterparts, which has been repeatedly reported by field inspectors and engineering teams. In this study, the dynamics of these pipelines which are at a higher risk of excessive stress and vibrational resonance during the passage of an ILI tool is investigated. 2D and 3D FE models are developed based on the Timoshenko beam theory to cover straight and curved geometries, and different boundary and loading conditions. The models predict the responses in the time and frequency domains which are used for the stress and frequency response analyses, and then to predict the risk of vibrational resonance during the passage of an ILI tool through a pipe segment. The proposed models are then verified experimentally and through FE analysis in ABAQUS software. The sensitivity analysis of the influential parameters is then carried out. Contrary to the conventional FE simulation approaches, the proposed model allows immediate response to a variety of typical and non-typical boundary conditions only at a fraction of the cost and time of the commercial FE software.To improve the robustness of the model, the application of strain sensors is also investigated. However, instead of utilizing costly, single point, strain sensors, a wide-surface piezoresistive nanocomposite sensor is proposed to be installed on critical spots along a line. The wide-surface nanocomposite sensor is used for collecting some boundary voltages when the sensor is stimulated with a known electric current. The voltages are then used for the prediction of the conductivity distribution across the sensor through machine learning (ML)-based electrical impedance tomography (EIT). The conductivity distribution across the piezoresistive nanocomposite sensor can finally be correlated with the strain distribution across the pipe.