Skip to main content
en

Moving boundaries: should the mesh deform or slide?

A valve opens, a piston travels, a flap folds. There are several ways to put motion into the model, and the choice depends on how much the geometry changes.

If a rotating component is involved, the path is clear — see the note on rotating machinery. This note is for motion other than rotation: a valve opening, a piston travelling, a flap folding, a body moving freely. The question here is what the motion does to the mesh.

Three approaches

Mesh deformation (smoothing). Nodes move with the boundary and the topology stays the same. For small, smooth motion this is the cheapest route. Its limit is clear: as cells stretch, quality falls, skewness rises and the solution degrades. If the motion exceeds a few times the characteristic cell size, deformation alone will not do.

Remeshing. Cells whose quality has degraded are regenerated during the run. It handles large motion, but the field is mapped onto a new mesh at every remesh and that mapping introduces numerical diffusion. Conservation error accumulates in a solution that remeshes frequently; monitoring the mass balance throughout the run is essential.

Overset (chimera) mesh. The moving component carries its own mesh and floats over a background mesh, with interpolation between them. For large and complex motion — a part separating, two bodies approaching — it does what neither deformation nor remeshing can. The price is that both meshes must be fine enough in the overlap region, and conservation is not guaranteed automatically.

The closing-gap problem

When a valve approaches full closure, or two surfaces touch, the volume between them goes to zero. No mesh can resolve that: as the cell size goes to zero the time step is forced to zero with it.

The practical remedy is to leave a small residual gap, typically on the order of one thousandth of the stroke. This is an assumption and it directly affects the leakage flow; its value belongs in the report, and where possible the case should be run at two different gaps to show how much the result moves.

The time step follows from the motion

The Courant number computed for the fluid is not enough here. There is a second condition: the boundary must not travel more than one cell in one time step. In a fast-closing valve this, not the flow, is usually the governing constraint.

Free motion needs two-way coupling

If the motion is known in advance — a piston stroke, a prescribed valve profile — the motion is imposed and the problem is one-way. But if the body moves under the flow's own force — a check valve disc, a separating part, a fluttering blade — the flow determines the force, the force determines the motion, and the motion changes the flow. That loop has to be closed with inner iterations at every time step; left open, the solution either drifts or diverges.

References

  1. Ferziger, J. H., Perić, M. & Street, R. L. (2020). Computational Methods for Fluid Dynamics, 4th ed. Springer. DOI: 10.1007/978-3-319-99693-6
  2. International Towing Tank Conference. ITTC Recommended Procedures and Guidelines 7.5-03-02-03: Practical Guidelines for Ship CFD Applications. ITTC.

Let's discuss your project

Let us pin down together the system to be analysed and the question that needs answering.

Let's discuss your project