Advancing two-phase flow Simulations: LSTM-Enhanced Reduced Order Modeling in CFD-DEM Applications
Please login to view abstract download link
This study introduces a novel approach in the fluid dynamics (CFD) and discrete element method (DEM) simulations field specifically focusing on fluid-solid systems. We propose a reduced order model (ROM) that utilizes Long Short-Term Memory (LSTM) networks to enhance the prediction and understanding of interactions between the two phases. The main goal of this model is to tackle the challenges and accuracy requirements when simulating two-phase flow systems. By combining LSTM with ROM, we aim to offer an accurate tool for capturing and predicting the dynamic behavior of both the Eulerian and Lagrangian phases. We evaluate the accuracy of our proposed ROM by comparing it with results obtained using the FOM for both Eulerian (volume fraction) and Lagrangian (particle position) variables in a fluidized bed system. Our findings demonstrate improvements in prediction accuracy as well as computational efficiency compared to the existing methods in the literature.
