The picture in the manual tells you the sequence of the nodes, not the node numbers themselves. If you look at your inp-file for the element, you will probably see that the element specification has the nodes not in ascending order.

Dear Martin,

Thank you for the response.

Yes, I agree that the picture in the manual (isoparametric node sequence) may indicates the sequence of the node and not the node number but if we look at the order of sequence between the manual and the particular elements are different. This discrepancy induces the inconsistency in the one to one mapping between the local and global node sequence while using the abaqus formulated isoparametric shape function.

Local node numbering sequence : 1, 2 ,3, 4, 5, 6, 7, 8

Global node numbering sequence: 5, 6, 8, 7, 1, 2, 4, 3.

For structured mesh, it is very ease to map the global node numbering sequence to the local node numbering sequence for instance, in this case NODE5 mapped to 1 , NODE 6 mapped into 2 NODE 8 mapped into 3 and so on .

But for the unstructured mesh how do map in a appropriate sequence inorder to preserve the abaqus formulated shape function based on their local coordinates ?

I don't understand what you are asking. The sequence of the node numbers is written to the inp-file in the element definition. You can also access it using the element connectivity in the python interface if you prefer that.

What exactly are you trying to do?

PS: If you use the mention-function, I'll get an alert...

Dear Martin,

Thank you for the reply.

I am writing VAM based finite element frame work numerical code for finding homogenized mechanical properties and local fields of composite materials. For which, I taking the composite meshed geometry details from ABAQUS. In this case mesh details are extracted by using PYTHON code but the problem is that while using the same sequence of element connectivity to calculate the stiffness matrix of each element through abaqus mentioned interpolation function, I am getting different solution. This issue I have been found due to the problem of ordering in the nodal connectivity in each elements. But, it can be solved by remapping element connectivity according to the abaqus given local node numbering sequence system (this may arise due to presence of determinant in some node mapping might be zero). With this remapping, I may able to get correct solution.

Particularly, this problem arises when the structured is discretized with sweep mesh (tapered element rather than regular shape elements ) rather than squared brick elements in abaqus. Martin Baeker

If the built-in connectivity routines do not allow to read this out correctly, you can always parse the inp-file with python. However, be aware that node numbers in an odb and the inp may differ...

The pattern of element connectivity varies from element to element for instance when I use sweep mesh algorithm in abaqus, in some element, connectivity of nodes in XY plane is on negative Z axis and then it is projected the consistent connectivity on positive Z axis, while for the other element nodal connectivity in XZ plane on negative Y axis is projected onto positive Y axis . This inconsistency provide the pattern of nodal connectivity in each element different when I use sweep mesh algorithm.

Secondly, I am extracting mesh details from ODB files rather than input file (but looking at mesh details on both ODB and Input file, they seems to be same in terms of node label and connectivity). Could you please suggest me that is there any option to control pattern of element connectivity in abaqus or some other option.

Thank you Martin Baeker

Dear martin,

I have written a PYTHON code which can extract and maintain the sequence of element connectivity individually of extracted HEXAHEDRON each elements but that individual element connectivity sequence which may not identical to the adjacent element connectivity for instance, one element, it may have the element connectivity sequence in clockwise and other adjacent element connectivity, it might be counter clockwise. The direction of element connectivity is seems to be difficult to be fixed when it comes to unstructured mesh case because I have done it for the case of structured mesh, while for the unstructured mesh it is difficult to be maintained the consistency. Could you please help me on this Martin Baeker

It will be really helpful Martin Baeker

Thank you

pandi

As I said, the sequence in the inp-file will always be consistent and according to the manual, so if nothing else works, use that. Although I find it surprising that the python routines in Abaqus should not be consistent (are you sure?).

Dear Martin,

Even looking at the input file where I found some element connectivity are not in consistent as mentioned in the manual. For your reference, I just highlighted in black bold lines:

Element connectivity:

1, 4, 33, 1, 21, 9, 34, 7, 8

2, 3, 19, 2, 33, 5, 22, 6, 34

3, 1, 33, 2, 18, 7, 34, 6, 23

4, 3, 33, 4, 20, 5, 34, 9, 24

5, 18, 23, 6, 2, 26, 25, 11, 12 (correct consistent is : 2,12,26,18,6,11,25,23)

6, 7, 23, 18, 1, 10, 25, 26, 13 (correct consistent is :1,8,26,13,7,23,25,10)

7, 3, 19, 22, 5, 15, 28, 27, 14

8, 19, 2, 6, 22, 28, 12, 11, 27

9, 20, 24, 9, 4, 29, 30, 17, 16

10, 5, 24, 20, 3, 14, 30, 29, 15

11, 21, 8, 7, 1, 31, 32, 10, 13

12, 9, 8, 21, 4, 17, 32, 31, 16

Thank you Martin Baeker

What do you mean by "consistent"? I assume that the ordering ABAQUS has in the inp-file is correct in the sense that the nodes (if looked at in the given sequence) for a hex with a positive volume and node ordering as in the manual, right? If not, that would be a case for the support...

Which node you chose as the first node is of course arbitrary.

For example, Element 1 to Element 4, the order of element connectives are projected from positive Z axis to negative z axis in counter clock wise fashion.

For your reference, written element 1 connectivity pattern and coordinates are :

Element1: 4 (273.50,-273.50,-50.0), 33 (0,0,-50.0), 1(-273.50,-273.50,-50.0) , 21(0,-386.78,-50.0), 9(273.50,-273.50,50.0), 34 (0,0,50.0,), 7(-273.50,-273.50,50.0), 8(0,-386.78,50.0)

But looking at the other elements connectives pattern for example let say 5, 6, 7 are in the same counterclockwise fashion but they are projected from positive X axis to negative X axis:

For reference, here I have written the 5 element connectivity and their coordinates are:

Element 5: 18 (-386.78,0,-50.0), 23(-386.78,0,50.0), 6(-273.50,273.50,50.0), 2(-273.50,273.50,-50.0), 26(-500,0,-50.0), 25(-500,0,50.0), 11(-500,500,50.0), 12(-500,500,-50.0)

Thank you Martin Baeker

Yes, there is no reason for abaqus to make sure that all nodes are in the same sequence with respect to some external coordinate system, it is of course totally fine for Abaqus to have one hex element labeld starting with the node in the lower north-east corner and the next starting in the upper south-west. In a general 3D component, element faces will not be aligned to any coordinate planes anyway.

If you want a mesh that is consistent in this way (for whatever reason...), you have to create it in some other software or (if it is very simple) using the NGEN/ELGEN features of the Abaqus inp code.

Thank you for the response.

Kindly let me know how to use those option (NGEN/ELGEN ) in ABQUS GUI so that I can extract the desired element connectivity pattern of mesh from ODB file via my PYTHON code .

Thank you Martin Baeker

"Kindly let me know how to use those option (NGEN/ELGEN ) in ABQUS GUI s"

You can't. Those are the basic keyword commands in an inp file to create a mesh.

You can read the connectivity of each element directly from the odb with python or you can simply parse the inp-file.

You can read the connectivity of each element directly from the odb with python : Yes, this is how I had extracted element via python but I need to modify same pattern projection of element connectivity . Martin Baeker

Hello Pandi,

The pattern for the nodes numbering as you have described above might not result a minimum bandwidth for the global sparse matrix of your system.

Please refere to Cuthill and McKee paper 1969 for more details about mesh numbering techniques or maybe what abaqus uses for its mesh numbering technique.

It seems that Abaqus will not use "the same pattern projection which you claim to be modified" since it will probably result in higher bandwidth for your graph.

Good question.. Please share me the best answer might you trust...

Regards…