I have implemented an ALE formulation into a 2D Euler solver, which Roe's Flux difference splitting and a ghost cell method for enforcing boundary conditions. I have found that when solid boundaries at moved at certain rates the solution fails (due to negative pressure developing in some cells). I would be grateful if anyone could tell me why this might be occurring. Thanks for your help

Neil

Do you update the geometry and re-distribute the cells in every time-step? That is needed for interior cells for each time-step. The other thing that I can think of is when the boudnary is moving very rapidly and the deformation happens in more than one layer of the grids, in which case (for free-surafec for example) is best to keep the changes within one layer first and get the simulation running. This avoids the mesh crossing in sharp gradinets. If you think this is any close to what you think may cause the problem then we can discuss it in more detail.

The boundary geometry is updated after each time-step and laplacian smoothing is preformed to redistribute the interior points. At the point of failure of the solution the grid quality is good, so this does not seem to be the problem. Also the deformationis such that the boundary does not cross the first layer

Explain your problem more. Are you solving for a phase change such as solidification, why would the solid boundary move in reality? The solution you gets may be unrealistic because the boundary conditions are unrealistic.

I am attempting to model the simple case of aerofoil undergoing forced oscillations of a sinusoidal nature. I found that as the amplitude of the oscilations was increased, but the period kept constant, there was a point at which the solution failed due to negative pressure, as explained earlier

Neil