Due to its characteristics, the semi-resolved approach requires a mesh that has cell sizes several times smaller than the characteristic length of the body that is immersed in the fluid. This is crucial for the correct characterization of flow using this method, since viscous (shear) and pressure (normal) forces need to be accurately obtained in the surface, in order to be able to predict the influence of the fluid on the particle and its velocity.
The method is called semi-resolved due to the fact that the flow field is evaluated at the surface, allowing for more precise prediction of the forces that the fluid exerts on the particle. The semi-resolved method should be used over the unresolved method when the following conditions are met:
The particles are at least an order of magnitude bigger than the cell sizes.
There is a need to accurately predict the forces acting on the particles.
The effects of the particles on the fluid flow are not negligible.
The shape of the particle is relevant for the interaction with the fluid flow.
The particles are flexible or have moving parts.
On the other hand, the unresolved method does not require a mesh so refined as the semi-resolved, since the fluid forces that act on the particles are obtained through the use of correlations and are implemented as source terms in the conservation equations. In the unresolved method, the mesh should have cells larger than the particles.
However, the accuracy of the solution obtained through the unresolved method depends heavily on the models utilized. If, for example, the shape of the particle has not been taken into account during the development of the selected fluid-particle interaction model, such as the drag correlation, this error is going to be propagated to the solution obtained. Also, the fluid forces influencing the particle are implemented as acting on its centroid, impossibilitating gradient of forces across the particles to be evaluated. Therefore, situations where the unresolved method is more appropriate involve the following conditions:
The mesh is relatively coarse when compared to the size of the particle (cells should ideally be an order of magnitude bigger).
Particles have more general shapes (such as spheres), which are usually covered under the most common interfacial force models.
The gradient of forces acting across the particle surface can be neglected.
Finally, a factor that should be considered when selecting the methodology to be used is the computational power available. Usually the requirement of high resolution meshes on the semi-resolved method is going to lead to a higher computational cost. However, the use of remeshing technique can be considered as an alternative to allow for more affordable solutions. With it, the mesh can be refined only at the surface of the particle. Then, for every timestep, the mesh can be recalculated, allowing for a coarser mesh on the domain in general, which can reduce heavily the time required to complete a simulation.