Doubts on restart with grid deformation

[rois1995]

I am trying to restart a simulation where a marker is moving. In particular, there is an airfoil that pitches and plunges.

I run the same simulation in three different ways:

1. One simulation from start to finish, no interruptions
2. Stop the simulation every n-th timestep and restart from the (n+1)-th timestep by using the initial mesh
3. Stop the simulation every n-th timestep and restart from the (n+1)-th timestep by using the mesh at the n-th timestep

The first two simulations lead to the same results. On the other hand, the third one deviates from the other 2 as the time increases. In the bottom images, there are the pressure and friction components of the vertical force on the airfoil (just for reference, also $F_x$ and $M_z$ are affected):

The more frequently I restart (thus the lower n is), the larger the discrepancies are.

I have noticed that the grid is being deformed differently between case 3 and the others, as shown in the following image after 30 time iterations. This difference in deformation appears as soon as the first restart is performed. Since the mesh deformation is performed with the Linear Elasticity Analogy, I assume that, when restarting, the element stiffness is different since in the first two cases it is computed over the starting grid, whereas in the third case it is computed on the deformed grid.

Are the discrepancies in results justified by the different grid deformation, or is there something else? Have you encountered something similar?

[pcarruscag]

I think it is because of the updated mesh, any other issue would also show up with the initial mesh.

[rois1995]

The stiffness matrix is computed with the updated coordinates or with the coordinates from the input grid?

[pcarruscag]

CMeshSolver always uses the initial coordinates. This makes it easier to differentiate transient cases with deforming grids.

[rois1995]

Ok, so restarting with the update mesh changes completely the ELA linear system, got it, thanks!