Gas-particle flows are commonly simulated through a two-fluid model at the industrial scale. However, these simulations need a very fine grid to have accurate flow predictions, which is prohibitively demanding in terms of computational resources. To circumvent this problem, the filtered two-fluid model has been developed, where the large-scale flow field is numerically resolved and small-scale fluctuations are accounted for through subgrid-scale modeling. In this study, we have performed fine-grid two-fluid simulations of dilute gas-particle flows in periodic domains and applied explicit filtering to generate data sets. Then, these data sets have been used to develop artificial neural network (ANN) models for closures such as the filtered drag force and solid phase stress for the filtered two-fluid model. The set of input variables for the subgrid drag force ANN model that has been found previously to work well for dense flow regimes is found to work as well for the dilute regime. In addition, we present a Galilean invariant tensor basis neural network (TBNN) model for the filtered solid phase stress, which can nicely capture the anisotropic nature of the solid phase stress arising from subgrid-scale velocity fluctuations. Finally, the predictions provided by this new TBNN model are compared to those obtained from a simple eddy-viscosity ANN model.

A novel sub-grid drag force model is proposed for coarse-grid Euler–Euler simulation of gas–solid fluidized beds. Starting from a transport equation for the drift velocity, an equilibrium condition is used as a basis to derive a new algebraic drift velocity model. The sub-grid correlations that show up are closed by a large-eddy PDF approach inspired from LES of turbulent reacting flows. The new analytical model only depends on the resolved slip velocity and on a few sub-grid moments of the solid volume fraction. Then, a conditional averaging procedure shows that the new model can be properly captured by a simple functional expression that only requires a closure for the sub-grid variance of the solid volume fraction. A priori validation studies show that the drift velocity is predicted with high accuracy (R ²>0.90) for a large range of filter widths and for both Geldart A and Geldart B particles.

A recent challenge in the modelling of particle flows is to build microstructure-informed drag models to overcome the average description of the fluid–particle force in the drag force correlations currently used in Euler–Lagrange and Euler–Euler models. To that end, we study through particle-resolved direct numerical simulations (PR-DNS) the flow past random assemblies of mono-dispersed spherical particles at three particle Reynolds numbers (10, 50, 100) and four solid volume fractions (0.10, 0.20, 0.30, 0.40). The present methodology is validated against theoretical and numerical results for the mean drag force in Stokes flows and finite Reynolds numbers flows. PR-DNS results are then used to characterize in details the statistics of the force distribution over the particle array, highlighting the substantial dispersion of the fluid force along the streamwise and transverse directions. The microstructure formed by the solid phase is described by means of a limited number of tensor quantities inspired from the fabric tensor used in granular media. Significant correlations are identified between the force experienced by a given particle immersed in a random array and a few key quantities that describe the anisotropy of its neighbourhood. A microstructure-based multi-linear model is proposed and validated against independent test cases. The model appears to perform best in the viscous and dense regimes. The addition of a stochastic contribution to the model allows to recover the correct level of force fluctuations at the cost of a lower correlation between the model and the data.