Predictive Study of Pressure Distribution in Gas-solid Fluidized Bed Flow Field Based on POD and TFM

The rapid prediction of pressure distribution in gas-solid fluidized beds is crucial for their operation monitoring and optimization. This paper presents an in-depth study of the rapid prediction method of pressure distribution in gas-solid fluidized beds. First, numerical simulations of the fluidized bed were performed using the two-fluid model (TFM) to obtain a variable-case dataset of the pressure distribution in the flow field inside the bed. The obtained simulation dataset was analyzed using proper orthogonal decomposition (POD) to extract eigenvectors and corresponding coefficients. Support vector machine regression (SVR) was then used to fit the projection coefficients, allowing the prediction of time-averaged pressure distribution based on feature vectors for gas-solid fluidized beds. The results indicate that predicting a single case takes only 0.04 CPU hours, significantly less than the 332 CPU hours required for simulation; the prediction results of the average bed pressure drop are highly consistent with the simulation results, and the mean absolute percentage error is within 0.01%. Therefore, POD can accurately and quickly predict the pressure distribution of fluidized beds. This paper examines the accuracy and validity of using POD in the gas-solid multiphase flow field. It provides a new research method and tool for optimizing fluidized bed engineering applications. © 2024 Science Press. All rights reserved.