Prediction of bend-pipeline erosion field based on CFD and reduced order method
Lei Mu, Shiyang He, Sheng Chen, Xizhong Chen, Jinhai SHI, Yongmin Zhang, Cenfan Liu
Abstract
In this paper the accuracy and efficiency of the Reduced Order Model (ROM), a machine learning algorithm, for predicting hydrodynamic behavior and erosion wear phenomena in a gas-solid two-phase flow pipeline system were investigated by performing a comprehensive comparison between ROM predictions and Computational Fluid Dynamics (CFD) results. Firstly, the accuracy of CFD simulation results among three erosion models, Oka, Vieira and DNV erosion models, was validated by comparison with experimental data. The Oka model was found to be more precise in predicting the erosion phenomena of the multiphase flow in pipeline. Secondly, the rational and efficiency of ROM predictions were examined, which the tendencies of hydrodynamic multiphase flow behaviors and erosion wear phenomena were discerned in agreement with CFD results. Furthermore, it is found that ROM predictions could save computational time by more than three orders of magnitude compared to the CFD method. Thirdly, the influence of design points and mode number on static pressure and erosion rate distribution was studied to enhance the accuracy of ROM predictions. Then, for a comprehensive validation of ROM predictions, four operating conditions with different gas inlet velocities and particle mass fluxes were analyzed. The result indicates that ROM predictions, under the recommended selection of design points and mode number, demonstrate a relatively good correspondence with CFD results both quantitatively and qualitatively. Finally, ROM predictions were utilized on a 135-degree industry pipe, enabling prompt and precise prediction of both the erosion position and erosion rate for the gas-particle industrial elbow pipe by conducting a comparison with the industrial measurement data.