Use Nodal Loads in Phase Change testcase

[Rich_B]

Hello,

Attached is a test of 'Use Nodal Loads = Logical True', using the test case 'PhaseChange3'. The sif files are lightly modified from the test case, such as from .ep to .vtu, and adding 'Calculate Loads = True' to both cases. Basically, adding 'Use Nodal Loads' definitely has an effect on the solution.

Rich.

[Rich_B]

one more observation, increasing the size of the time step in both cases causes the mesh update to explode. For example, using the original test case, a time step of 6.0e-2 runs 100 steps okay, while a time step of 6.5e-2 runs okay at first, then oscillates before exploding. Are there settings that would help increase the time step and still retain stability?

Rich.

[kevinarden]

No not really, mesh updater is intended for small displacement changes. What is happening is that the displacement are so large, the elements are too distorted to form a stiffness matrix from. There is no time stepping method to over come this. It can be demonstrated by reducing the displacement of the mesh

Mesh Update 2 = Variable "PhaseSurface"
Real MATC "tx(0)*0.0001"

and the problem runs to completion at the high time step. You can then scale the results in paraview and you can see the amount of deflection and element distortion that is being created. If I scale the mesh update results to full scale, the nodes are being moved 3.1 meters. The only way to over come this with actual displacements is to adapt the mesh in mesh update based on the element distortion.

displacement.png (35.29 KiB) Viewed 259 times

[Rich_B]

Hello Kevin,

Thanks for checking the test cases. That makes perfect sense about the magnitude of the displacements and scaling down the displacement to prove the issue is brilliant.

Rich.