Hello:

I have read some quotes on meshing schemes and corressponding problems and accuracy.

However, I have observed in my studies that for pure tetrahedral meshes, there exists a threshold beyond which the solution will be further away from the analytical value than the previous or coarser mesh.

1. Do you think that is possible?

2. do you think that this happens due to the fact that numerical error for tetrahedral cells increases exponentially leading to a condition that any further discretization will only take you away from the analytical solution? (in other words the numerical error due to the tetrahedral cells is much higher than the improvement in the solution due to further discretization in the form more tet cells)

Kindly reply and thank you for yout time and thoughts.

Best regards Sa

Sam,

Tetrahedral meshes are among the least accurate type of meshes, but they should still provide you more accurate solutions with mesh refinement. Can you be sure there is no bug in your code? What case are you running? I have experienced some problem where locally you cannot achieve mesh independence because the solution is singular (ie. sharp corners, cracks, etc.). However, the overall accuracy shall improve with mesh refinement. If I were you I would check two things: 1. Refine the mesh further and check what is happening. The error shall start decreasing after some point unless there is a bug in the code. 2. Search for a bug in my code.