hexahedral mesh not computable

Alle Fragen zu: Vernetzung, Materialien, Lasten, Randbedingungen und Elementparameter /
All questions to: meshing, materials, boundary conditions and element properties

Moderatoren: ccad, mz15, auroraIco, Lehrstuhl

Antworten
phb3
Newbie
Beiträge: 7
Registriert: Mo 16. Nov 2015, 19:04

hexahedral mesh not computable

Beitrag von phb3 »

Dear Z88 Users,

I'm testing Z88, and generated a mesh of hexahedral no. 1 elements using Microsoft Excel.
The mesh passes the z88vry.pl test, but is not computable.
The z88det.txt file identifies all elements.

I believe that the node order is correct to generate a right-handed local coordinate system, compatible with the description in the theory manual.

The Aurora structure file is linked at: http://1drv.ms/1Od7kk9

I would very much appreciate your experienced eyes on the file.
Danke für Ihre Hilfe und Aufmerksamkeit!
ClaKle
Alumni
Beiträge: 36
Registriert: Mi 3. Dez 2014, 13:20

Re: hexahedral mesh not computable

Beitrag von ClaKle »

Hi phb3,

the order of the nodes in your structure is wrong. Just switch the first 4 nodes with the last 4.

Examle for the first element:
1 1
13 65 66 14 1 53 54 2

I have tried this with the first elements and this approach solved the problem.

I hope this answer will help you.

Kind regards,
ClaKle
phb3
Newbie
Beiträge: 7
Registriert: Mo 16. Nov 2015, 19:04

Re: hexahedral mesh not computable

Beitrag von phb3 »

Thanks, ClaKle - I fear that I may be missing a basic concept.

Procedure:
start a new Z88 Aurora project
import a z88 text file
select preprocessor
select Super elements
select Mesh info

Result: mesh not computable alert.

I get the same result with my original node order, or the swapped node order of your kind suggestion.

I appreciate your time and effort reviewing this work.

original order

3 8 1 24 0
1 3 0.000500 0.000000 0.000000
2 3 0.000433 0.000250 0.000000
3 3 0.000373 0.000000 0.000000
4 3 0.000323 0.000186 0.000000
5 3 0.000500 0.000000 0.000500
6 3 0.000433 0.000250 0.000500
7 3 0.000373 0.000000 0.000500
8 3 0.000323 0.000186 0.000500
1 1
1 2 4 3 5 6 8 7


swapped order:

3 8 1 24 0
1 3 0.000500 0.000000 0.000000
2 3 0.000433 0.000250 0.000000
3 3 0.000373 0.000000 0.000000
4 3 0.000323 0.000186 0.000000
5 3 0.000500 0.000000 0.000500
6 3 0.000433 0.000250 0.000500
7 3 0.000373 0.000000 0.000500
8 3 0.000323 0.000186 0.000500
1 1
5 6 8 7 1 2 4 3
Dateianhänge
basic structure.jpg
element geometry, identify node order
(60.33 KiB) 1139-mal heruntergeladen
ClaKle
Alumni
Beiträge: 36
Registriert: Mi 3. Dez 2014, 13:20

Re: hexahedral mesh not computable

Beitrag von ClaKle »

Hi phb3,

In z88r.log you find detailed information why the project is not computable. There is said that the determinant is less than/equal 0. The reason for this message is probably the structure of the mesh. In a few elements in the middle you see big angles near 180°. In the following picture I have marked the area. These elements are to strong distorded.
prob.jpg
prob.jpg (47.4 KiB) 2943 mal betrachtet
Two possibilities to solve your Problem:
1. Change the position of the kritical nodes (40, 42, 44, 46, 48, 92, 94, 96, 98, 100) to make the angle smaller.
2. Add a 4th, 5th, 6th.... ring of elements in the middle and keep a very small gap at the center.

Kind regards,
ClaKle
phb3
Newbie
Beiträge: 7
Registriert: Mo 16. Nov 2015, 19:04

Re: hexahedral mesh not computable

Beitrag von phb3 »

Hi ClaKle,

Thanks so much for mentioning z88r.log. I cannot find this file on my system.
Perhaps this missing file is a clue to my error.

In the project directory, the files are
project.z88
z88control.txt
z88det.txt
z88i1.txt
z88nc.log
z88structure.txt
z88trail.txt

In C:\Z88AuroraV3\win\bin, there are four recently modified files:
z88aurora.log
z88enviro.dyn
z88.fcd
z88.dyn

I appreciate your advice about shallow angles. However the error persists in the single element, 8 node test case which I submitted.

Since the b27 example file trommel.txt imports as a computable mesh, I subselected the data for the first element, which also imported as a computable mesh.

One thing which does result in a computable mesh is to multiply the coordinates in one of my single element test files by 1000, for example:
3 8 1 24 0
1 3 0.500 0.000 0.000
2 3 0.433 0.250 0.000
3 3 0.373 0.000 0.000
4 3 0.323 0.186 0.000
5 3 0.500 0.000 0.500
6 3 0.433 0.250 0.500
7 3 0.373 0.000 0.500
8 3 0.323 0.186 0.500
1 1
1 5 7 3 2 6 8 4

Are the units of measurement (eg millimeters vs meters) an important consideration in using Z88 Aurora ?

Best regards,

phb3
ClaKle
Alumni
Beiträge: 36
Registriert: Mi 3. Dez 2014, 13:20

Re: hexahedral mesh not computable

Beitrag von ClaKle »

Hi phb3,

The file z88r.log only exists when you have started the solver. I guess you have just checked the mesh info, so you don´t get this file. So this is not the problem.

After a closer look on your z88structure.txt, I recognized that the numerical values of your coordinates are very small (e.g. 6.12574E-20). The query limit of the determinant is about 10E-8. So you have to use bigger values.

Concerning units: In principle it doesn´t matter which units you choose, as long as you concider the consistence. Normally in FEA you use mm, t, N and derive all other units, e.g. N/mm².

Best regards
ClaKle
Antworten