Description
Industrial scale processes feature billions of particles and wide particle size distributions. EulerLagrange (EL) simulations of such processes, if at all possible, demand very high computational resources, especially for polydisperse cohesive (wet) gas-particle systems. As a result, it is necessary that the number of tracked particles is reduced with some sort of coarse-graining (CG) [1,2]. Crucial for the imitation of a primary particle system with computational parcels is the correct mapping of the coarsened Lagrangian information onto the Eulerian Grid. Furthermore, it is well known that the fluid grid resolution has a significant impact on aerated powder flow predictions, e.g., the voidage distribution is decisive for the prediction of the drag force [3]. Therefore, a smoothing operation that distributes the exchange fields to the surrounding cells is necessary, since simulations become numerically unstable if the particle size exceeds the fluid cell size. In polydisperse systems the fluid cell size must be based on the smallest particles size to achieve an accurate prediction of mesoscale flow features [4]. Typically, the bigger particles are several times larger than the smallest particles, and subsequently exceed the size of the fluid cell even without coarsegraining. Consequently, a trade-off between (i) modelling wide particle size distributions and (ii) predicting flow features in great details has to be made. Our contribution first introduces an optimal smoothing operation for the flow prediction of monodisperse powders. Thereby the objective function was to recover the same domain-average slip velocity for different coarse-graining ratios in a fully periodic domain. It was found that the optimal strategy is to use a linear relation between the length of the smoothing filter and the parcel size when maintaining the same fluid cell size. Second, the optimal smoothing lengths for monodisperse systems were applied in a polydisperse system. Results for different smoothing lengths and fluid cell sizes are ultimately benchmarked against data from expensive simulations of the primary system. It is demonstrated that deviations smaller than 5% for a range of cohesive regimes and coarse-graining ratios can be achieved when using optimal smoothing lengths and a coarse-graining model based on constant stresses.| Period | 28 Oct 2019 |
|---|---|
| Event title | VI International Conference on Particle-Based Methods |
| Event type | Conference |
| Location | Barcelona, Spain |