'Tutorial - Bearingblock mesh' simulation project by FilippoMenconi


I created a new simulation project called 'Tutorial - Bearingblock mesh':

This meshing tutorial shows how to setup the Tetrahedral automatic, Tetrahedral parametric and the Tetrahedral with local refinements for a structural analysis simulation.

More of my public projects can be found here.


With a couple of bearingblocks we can locate an axis in the space. Thanks to its structure a bearingblock can locate a bundle of lines, so a bundle of axes; this way we can avoid the problems related to machining errors.


The mesh is made by two elements:

  • a set of coordinates who defines the position of the nodes
  • a law of connectivity, who defines which nodes form each element

The mesh has to correctly represent the physical domain, so it has to correctly represent the external boundaries and internal boundaries too (interfaces).

A point has to be either a node or part of an edge, there is not space for any "crisis of identity". It is also possible to work with grids that don't respect this property.


One of the key aspects of finite element analysis is the dimension of the elements. We can use the diameter of the sphere that circumscribes the element (if the dominion has three dimensions, the diameter of the circle that circumscribes the element if the dominion has two dimensions)to define the dimension of the elements. 

Smaller is h more accurate is the solution of the problem. An accurate solution costs very much, in terms of computational time.

The optimal choice for h depends on the solution of the problem, e.g. if there is an area characterized by small gradients of the solution we can use a coarse mesh, so it varies locally. The optimal mesh is an unknown of the problem.


When I speak about an optimal mesh I speak about a mesh that allows me to determine a solution arbitrarily close to the analytical one. The optimal mesh is a compromise between the gradients of the solution and the dimension of the elements.