# Direct numerical simulation of particulate flows with an overset grid method

## TL;DR

This paper presents an overset grid method for simulating particulate flows that accurately captures boundary layers at reduced computational cost, demonstrating improved efficiency and accuracy over traditional methods.

## Contribution

The authors develop and validate an overset grid approach for particulate flow simulation that enhances boundary layer resolution and computational efficiency compared to existing methods.

## Key findings

- Achieves accurate boundary layer resolution with fewer grid points.
- Reduces CPU time by a factor of 13 compared to uniform grid methods.
- Maintains accuracy while significantly decreasing computational cost.

## Abstract

We evaluate an efficient overset grid method for two-dimensional and three-dimensional particulate flows for small numbers of particles at finite Reynolds number. The rigid particles are discretised using moving overset grids overlaid on a Cartesian background grid. This allows for strongly-enforced boundary conditions and local grid refinement at particle surfaces, thereby accurately capturing the viscous boundary layer at modest computational cost. The incompressible Navier--Stokes equations are solved with a fractional-step scheme which is second-order-accurate in space and time, while the fluid--solid coupling is achieved with a partitioned approach including multiple sub-iterations to increase stability for light, rigid bodies. Through a series of benchmark studies we demonstrate the accuracy and efficiency of this approach compared to other boundary conformal and static grid methods in the literature. In particular, we find that fully resolving boundary layers at particle surfaces is crucial to obtain accurate solutions to many common test cases. With our approach we are able to compute accurate solutions using as little as one third the number of grid points as uniform grid computations in the literature. A detailed convergence study shows a 13-fold decrease in CPU time over a uniform grid test case whilst maintaining comparable solution accuracy.