Hello everybody,
I came here to share my experience and get some view from a side: maybe someone give good idea, suggestion will see my mistakes.
So, I am simulating compression of RVE: hyperelastic matrix (cube 70x70x70 mm with holes for inclusions) filled with spherical rigid (compared to the matrix) inclusions (diameter=8mm, total 181 inclusions, total volumic fraction of inclusions ~ 0.225).
Inclusions and matrix are modeled separately, meshed separately and after tie-constrained.
So my two main problems are meshing of the matrix and constraining.
About meshing:
The fraction of inclusions is quiet high, so, sometimes they are located very close to each over(if I am not mistaken the smallest distance is 0.06 mm). But some inclusions are located on good distance from overs (more than 1 mm). Automatic meshing in Abaqus is resulting in bad mesh (warning elements near inclusion/holes, which are "flat" - very small in one direction) or millions and millions of elements. So I tried to mesh matrix of my RVE in HyperMesh. It gave me really better result in meaning of elements quality and elements number (By the way in HyperMesh I used element sizes allowed from 2 mm down to 0.4 mm). Abaqus "Verify Mesh" tool doesn't show any warrings or errors.
However, during data check "2 elements are distorted" warring arises
About tie-constrain(surface-to-surface):
Before going to details, I need to say that I am using automatic contact pair identification, because I don't see possibility to do it by hands (181 pairs, everything inside the geometry).
So, HyperMesh result can be exported to Abaqus as an orphan mesh, so I am doing it. Here some problem start:
1) Abaqus can't successfully identify contact pairs between Orphan mesh(matrix) and meshed geometry(inclusions) - software creates pairs with only master or slave surface. To solve this I did an Orphane mesh from meshed inclusions. In the case of two Orphan "geometries" identifications goes better (still some adjustments are necessary, like flipping master/slave).
2) By default tie constraint make an adjustment of slave nodes to master surface. In manual it is advised to keep it like this. However, adjustment leads to "zero volume element" errors end terminating the calculation. I don't have a clue how to solve this (except of disabling of this adjusting). Any ideas?
In attachment I added screenshot from HyperMesh (so you can see location of holes inside the matrix)
I came here to share my experience and get some view from a side: maybe someone give good idea, suggestion will see my mistakes.
So, I am simulating compression of RVE: hyperelastic matrix (cube 70x70x70 mm with holes for inclusions) filled with spherical rigid (compared to the matrix) inclusions (diameter=8mm, total 181 inclusions, total volumic fraction of inclusions ~ 0.225).
Inclusions and matrix are modeled separately, meshed separately and after tie-constrained.
So my two main problems are meshing of the matrix and constraining.
About meshing:
The fraction of inclusions is quiet high, so, sometimes they are located very close to each over(if I am not mistaken the smallest distance is 0.06 mm). But some inclusions are located on good distance from overs (more than 1 mm). Automatic meshing in Abaqus is resulting in bad mesh (warning elements near inclusion/holes, which are "flat" - very small in one direction) or millions and millions of elements. So I tried to mesh matrix of my RVE in HyperMesh. It gave me really better result in meaning of elements quality and elements number (By the way in HyperMesh I used element sizes allowed from 2 mm down to 0.4 mm). Abaqus "Verify Mesh" tool doesn't show any warrings or errors.
However, during data check "2 elements are distorted" warring arises
About tie-constrain(surface-to-surface):
Before going to details, I need to say that I am using automatic contact pair identification, because I don't see possibility to do it by hands (181 pairs, everything inside the geometry).
So, HyperMesh result can be exported to Abaqus as an orphan mesh, so I am doing it. Here some problem start:
1) Abaqus can't successfully identify contact pairs between Orphan mesh(matrix) and meshed geometry(inclusions) - software creates pairs with only master or slave surface. To solve this I did an Orphane mesh from meshed inclusions. In the case of two Orphan "geometries" identifications goes better (still some adjustments are necessary, like flipping master/slave).
2) By default tie constraint make an adjustment of slave nodes to master surface. In manual it is advised to keep it like this. However, adjustment leads to "zero volume element" errors end terminating the calculation. I don't have a clue how to solve this (except of disabling of this adjusting). Any ideas?
In attachment I added screenshot from HyperMesh (so you can see location of holes inside the matrix)
