Discussion Forum

Hello Alaa,

In a 3D model, triangles, quads, edge elements and vertex elements are not counted under the "Domain element statistics" header because they do not pertain to domains (domains in the 3D context meaning volumes).

Jeff

Hello Alaa,

In a 3D model, triangles, quads, edge elements and vertex elements are not counted under the "Domain element statistics" header because they do not pertain to domains (domains in the 3D context meaning volumes).

Jeff

Hello Jeff,

Thank you for your reply. Now I understand that when I only sum Tetrahedra, Pyramids, and Prisms, I get the exact value of "Number of elements" below.

However I have another question, since I'm doing a mesh independence study, I'm trying to calculate the total number of elements (number of grids). Should I consider the "Number of elements" to be the total (which is 12 Million), or should I also add to that the Triangles, Quads, Edge elements, and Vertex elements (which is 13 Million) ?

Because I can't find an expression to plot only the "Number of elements" maxop1(meshelementall) > plots the total of 13 Million maxop1(meshelement) > plots only the Tetrahedra

Thanks.

Hello Alaa,

I wouldn't be too concerned about the difference. In practice, the point of a convergence study is to check whether your mesh is fine enough that results would remain essentially the same if you used a finer mesh. This can be more an art than a science. Strictly speaking, you cannot usually prove that the solution would not change if you could use an even finer mesh than the finest one your computer can handle. As is often the case in engineering, you should use common sense and engineering experience to assess whether there is any particular reason why the solution would vary if you were able to refine the mesh further.

Jeff

PS: In an academic setting, a convergence study aimed at determining the convergence rate of an element formulation is ordinarily performed against the element size. The number of domain elements varies with the cube of the element size whereas the number of surface elements varies with the square of the element size, so that asymptotically (which what matters in such a convergence study) the number of surface elements becomes negligible compared to the number of domain elements and you would therefore obtain the same convergence rate whether you plot the error aagainst the element size, the total number of elements or the number of domain elements. But note also that those types of studies, to be rigorous, require sequences of FEA spaces that are subspaces of each other - something that is not easily achieved other than with structured meshes.