Early Detection and Prevention of Two-Phase Flow-Induced Vibrations in CO2 Transport Pipelines Using a Flow Control System Coupled with a Neural Network Model

This study addresses the challenge of two-phase flow-induced vibrations (FIVs) in CO2 transport pipelines, which are crucial for safe operation of the pipelines and processing facilities for carbon capture and storage (CCS) projects. The formation of two-phase flow due to phase changes may induce vibrations of pipelines, especially in riser sections of offshore platforms. Conventional design strategies to mitigate FIVs may not be applicable to the CO2 transport pipelines in the case of reuse of the existing infrastructures. This study investigates the performance of a flow control system coupled with a machine learning algorithm for early detection and prevention of FIVs. Experiments were conducted using a flow loop using air and water to simulate two-phase flow and to obtain both vibration and pressure fluctuation data, and then those data were used to develop and train an artificial neural network (ANN) model. This ANN model effectively identifies two-phase unstable flows (USFs) and initiates an internal model control logic to adjust the valve opening of the flow control system. Upon adjusting the valve opening, stable two-phase flow was obtained. The obtained results demonstrated that early detection and timely control should reduce the maximum pressure and the duration of USFs. The control system was also applied to the simulation model for the offshore CO2 transport pipeline for the CCS project in the East Sea of Korea and proved its effectiveness to minimize the impact of two-phase USF. These results suggest that implementing this ML-based control system enhances the safety and reliability of CO2 transport pipelines, making CCS projects more economically viable by utilizing existing infrastructure.