Modeling of the filtered drag force in gas-solid flows via a deep learning approach

A deep learning approach based on a proposed artificial neural network (ANN) is exploited to predict the filtered drag force, which is a desirable term for coarse-grid simulations of gas-solid flows. The proposed ANN model includes one input layer, five hidden layers and one output layer. The hidden layers are composed of three convolutional layers and two fully-connected layers. The results show that the proposed ANN model is superior to the multi-layered perceptron (MLP) model and the best available traditional functional model over a wide range of the filter size. The results also indicate that the use of the information on neighboring coarse grids is necessary for accurate estimation of the filtered drag force at the considered filter sizes. This finding is substantiated by a budget analysis which demonstrates that the diffusion of the Reynolds stress and solid phase stress significantly affects the filtered drag force. (c) 2020 Elsevier Ltd. All rights reserved.