Research On Flow Pattern Identification Of Gas-Liquid Two Phase Flow In Subsea Jumper Based On Efficient Net And Electrical Capacitance Tomography

Subsea jumper is a pipeline connector that used to connect underwater manifold,Christmas tree,and pipeline terminations in subsea oil and gas production facilities.It is widely employed in deep-sea oil and gas development projects.A typical M-shaped rigid subsea jumper consists of multiple bends and long horizontal spans.Due to its unique structural configuration,the internal multiphase flow characteristics become complex and variable,making the pipeline susceptible to stress impacts and vibrations,leading to fatigue failure.Therefore,ensuring the stable operation of subsea jumpers is an urgent problem to be addressed.Flow pattern precisely represents the state of multiphase flow,and its identification can explain the variations of other flow parameters.The changes in the flow pattern of subsea jumpers directly impact their safe operation,hence it is necessary to conduct research on flow pattern identification and prediction.This thesis focuses on the typical M-shaped rigid subsea jumper as the research object.With the support of the Experimental Platform for Subsea Multiphase Flow at Dalian Maritime University,the research investigates the flow characteristics and flow pattern evolution of gas-liquid two-phase flow inside the subsea jumper under different gas void fractions,superficial velocities,and mixture velocities,using visualization experiments and Computational Fluid Dynamics.Firstly,an experimental platform for gas-liquid two-phase flow in the subsea jumper is established.Through experimental observations,it is found that the subsea jumper exhibits multiple flow patterns simultaneously due to its unique structural configuration.When slug flow occurs in the pipeline,the dynamic fluid forces generated impose a high impact load on the pipeline,resulting in structural vibrations.Additionally,based on the ANSYS 2020R1 FLUENT simulation software,numerical simulations are conducted to investigate the flow evolution of gas-liquid two-phase flow in the M-shaped rigid subsea jumper.The pressure fluctuations and flow pattern distribution when the gasliquid two-phase flow passes through bends and other areas are studied.Furthermore,combining the Electrical Capacitance Tomography technology,flow pattern data obtained from the observation of the horizontal section of the subsea jumper are reconstructed.A total of 3,271 images of four typical flow patterns,including bubbly flow,annular flow,slug flow,and stratified flow,are selected.Clear and distinct flow pattern samples are manually classified to establish a flow pattern identification dataset.Finally,using the Tensor Flow framework platform,the Efficient Net network model is built.The flow pattern dataset is trained and learned using the model.Through model comparison,it is demonstrated that the Efficient Net-B5 network model is the most suitable for this research,with a flow pattern identification accuracy of 95.6%.To improve the recognition accuracy of similar flow patterns,the SE attention mechanism is introduced,and the Adam optimizer is used to improve the network model,enhancing the feature extraction capability and classification accuracy.The optimized Efficient Net-B5 network model achieves an accuracy of 97.55% in flow pattern identification.The research results of this thesis are conducive to a deeper understanding of the transition rules of the flow pattern inside the jumper,and provide theoretical and practical guidance for the design and optimization of subsea jumpers in subsea oil and gas field development.It promotes the further development and application of multiphase flow characteristics and flow pattern recognition technology,and contributes to the sustainable development of subsea oil and gas field development.