It is the "Homogenized SG volume" that is wrong for the circular. I assume your diameter D=100, so then the homogenized SG volume should be its area A=(pi*D^2)/4. Now you look at the .sc file, last row, it is 10000, obviously, it calculate as D^2. I do not know if this is a bug by the GUI itself or user manually input this volume because I am not familiar with GUI. Please check change the volume to its area and see how results go.
As Ning said, the problem is omega (the last number). I did your cases again from the .cae file, and got the same E1=192GPa for both SGs.
By default, if you leave the 'Omega' field blank in the GUI, it will calculate the area by finding the bounding limits in two directions, as if the part is a rectangular. So you need to calculate the area by yourself if the part is not a rectangular.
It is hard to say that this is a bug, since this is what the 'Omega' input box is designed for. It is not easy to calculate the area of a shape other than rectangular automatically. Adding areas of all elements is not reliable since there may be voids, just like your cases. Thus for non-rectangular shapes, user needs to calculate the area by hand and put that number into the 'Omega' box in the 'Homogenization' dialog window.
Fantastic! Seems like the problem is solved. But yet here's another thing.
Now instead of isotropic material, I want to use an orthotropic material (E1 = 53.59, E2 = 11.72, E3 = 12.35, v12 = 0.471, v13 = 0.161, v23 = 0.294, G12 = 3.43, G13 = 3.15, G23 = 3.10) for both the circular and the square RVE having 4% hole for each of them at the center. For the case of circular RVE, I am giving input of the Omega value manually which is equal to the circular area. Now for the circular RVE, I am getting the E1 value that I should get, which is equals to 51.43 GPa. But the E2 and E3 values I am getting in case of circular RVE doesn't make sense. Because after homogenization, I am getting E2 = 1.96 GPa and E3 = 1.97 GPa. From square RVE I am getting E2 = 10.43 GPa, E3 = 10.95 GPa, which do make sense.
Could you please tell me why I am not getting a similar value for E2 and E3 while I am using circular RVE?
Thanks in advance for your help. Please accept my sincere appreciation for it.
The circular one with a hole inside it cannot be dealt with the current version of SwiftComp because of the existence of the hole. You can try to fill the hole with a very weak material. The result will be close to the result that you obtained from using uniform traction boundary conditions.
The circular one with a hole inside it cannot be dealt with the current version of SwiftComp because of the existence of the hole. You can try to fill the hole with a very weak material. The result will be close to the result that you obtained from using uniform traction boundary conditions.
I am using Abaqus-SwiftComp GUI to generate input file.
I rotate and translate the instance in assembly so that x,y,z axes and coordinate center is aligned to what I want. But after I generate the input file, it seems nodal coordinates are not from assembly, then I figured they are from part/instance.
1. Are you able to change code so that nodal coordinates are from assembly? Because sometimes we rotate and translate instance so that axes are aligned to what we want (picture: assembly.png).
instance.png : default coordinates of instance before I translate and rotate in assembly
assembly.png : coordinates after I translate and rotate
If you just want to perform homogenization to get effective properties, you can use TexGen4SC at https://cdmhub.org/tools/texgen4sc , which integrates TexGen and SwiftComp. I think the Abaqus-SwiftComp developers may be able to answer the problem you met, but I am not sure if you can get the right results from the GUI since the material local orientations are defined in an unusual way in the output files from TexGen.
Usually, I directly use TexGen4SC to perform homogenization for textile composites. There are a lot of unnecessary definitions of nodal sets, steps and BCs in the inp file from TexGen for SC-Abaqus GUI, but if you want to use python script from TexGen website, then all the definitions are needed. I am not sure if this caused your problem. Another possible issue is the model information is stored in different files (at least in two files .inp and .ori), I am not sure if SC-Abaqus GUI can find all information. I am not an expert for SC-Abaqus GUI, may be the GUI developers can give you more accurate answers.
Reza, if you export TexGen into Abaqus file, then I will expect you to use Abaqus for the RVE analysis because that file is created for that specific purpose. Otherwise, you should export it to SwiftComp file, then you can either use TexGen4SC directly. If you want to use Abaqus-SwiftComp GUI (I don’t understand why), then you should use the SC file generated from TexGen4SC.
Please understand Abaqus-SwiftComp GUI is for you to perform homogenization and dehomogenization using a model created by Abaqus or a SC file created by other preprocessor such as TexGen or Gmsh. Abaqus-SwiftComp GUI does not perform macroscopic analysis. The properties out of homogenization should be used as inptus for regular Abaqus analysis.
My goal is to homogenize a unit cell which represents a SG of a woven fabric.
Then, use the properties to analyze a woven fabric (which I want to model as a 2D shell) under moment applied to one edge while the other edge is clamped.
Next, dehomogenize it to capture the local fields, stress and displacement fields, at any point of the structure.
The second step will be done using Abaqus. Abaqus-SwiftComp GUI will do the first (homogenization) and last step (dehomogenization).
---- Emailed forum response from wenbinyu@purdue.edu
I am trying to homogenize my model (unit cell imported from TexGen) in Abaqus-SC, but I assume homogenization is not carried out since I cannot import homogenized properties.
"Inputs echoed in file UnitCell_nSG3_2D_C3D6pbc.sc.ech!
Constitutive modeling for a 2D (plate/shell) model!
Homogenization will be carried out!
You are running SwiftComp Parallel Version!"
I would appreciate it if you could help me address the problem.
Ning Liu @ on
Can you provide direct input files? Files ending in .sc
If you can also find .ech .k, please attach as well
Anonymous @ on — Edited @ on
Please go to the link below and find files named "Files for Fig. 1" and "Files for Fig. 2"
https://drive.google.com/open?id=0BzDQpu7yML7dUnEyOU9tMEhhZkE
Please let me know if you can't open the folders or files.
Thanks for your generosity.
Ning Liu @ on
Hi,
It is the "Homogenized SG volume" that is wrong for the circular. I assume your diameter D=100, so then the homogenized SG volume should be its area A=(pi*D^2)/4. Now you look at the .sc file, last row, it is 10000, obviously, it calculate as D^2. I do not know if this is a bug by the GUI itself or user manually input this volume because I am not familiar with GUI. Please check change the volume to its area and see how results go.
Su Tian @ on
Hi,
As Ning said, the problem is omega (the last number). I did your cases again from the .cae file, and got the same E1=192GPa for both SGs.
By default, if you leave the 'Omega' field blank in the GUI, it will calculate the area by finding the bounding limits in two directions, as if the part is a rectangular. So you need to calculate the area by yourself if the part is not a rectangular.
Anonymous @ on
Thanks for the reply. Isn't it possible to fix it using GUI without using inp file?
Su Tian @ on
It is hard to say that this is a bug, since this is what the 'Omega' input box is designed for. It is not easy to calculate the area of a shape other than rectangular automatically. Adding areas of all elements is not reliable since there may be voids, just like your cases. Thus for non-rectangular shapes, user needs to calculate the area by hand and put that number into the 'Omega' box in the 'Homogenization' dialog window.
Anonymous @ on — Edited @ on
Fantastic! Seems like the problem is solved. But yet here's another thing.
Now instead of isotropic material, I want to use an orthotropic material (E1 = 53.59, E2 = 11.72, E3 = 12.35, v12 = 0.471, v13 = 0.161, v23 = 0.294, G12 = 3.43, G13 = 3.15, G23 = 3.10) for both the circular and the square RVE having 4% hole for each of them at the center. For the case of circular RVE, I am giving input of the Omega value manually which is equal to the circular area. Now for the circular RVE, I am getting the E1 value that I should get, which is equals to 51.43 GPa. But the E2 and E3 values I am getting in case of circular RVE doesn't make sense. Because after homogenization, I am getting E2 = 1.96 GPa and E3 = 1.97 GPa. From square RVE I am getting E2 = 10.43 GPa, E3 = 10.95 GPa, which do make sense.
Could you please tell me why I am not getting a similar value for E2 and E3 while I am using circular RVE?
Thanks in advance for your help. Please accept my sincere appreciation for it.
Bo Peng @ on
The circular one with a hole inside it cannot be dealt with the current version of SwiftComp because of the existence of the hole. You can try to fill the hole with a very weak material. The result will be close to the result that you obtained from using uniform traction boundary conditions.
Bo Peng @ on
The circular one with a hole inside it cannot be dealt with the current version of SwiftComp because of the existence of the hole. You can try to fill the hole with a very weak material. The result will be close to the result that you obtained from using uniform traction boundary conditions.
Ning Liu @ on — Edited @ on
Hello Bo and Su,
I am using Abaqus-SwiftComp GUI to generate input file.
I rotate and translate the instance in assembly so that x,y,z axes and coordinate center is aligned to what I want. But after I generate the input file, it seems nodal coordinates are not from assembly, then I figured they are from part/instance.
1. Are you able to change code so that nodal coordinates are from assembly? Because sometimes we rotate and translate instance so that axes are aligned to what we want (picture: assembly.png).
instance.png : default coordinates of instance before I translate and rotate in assembly
assembly.png : coordinates after I translate and rotate
Reza Sourki @ on
Hi all,
I imported a model from TexGen into Abaqus-SwiftComp. I received the following error while trying to import the homogenized properties.
File "C:\User\...\abaqus-swiftcomp-gui\scMacroMat.py", line 238, in importSCmat
stiffnessMatrix=prop_matrix_tuple,
UnboundLocalError: local variable 'prop_matrix_tuple' referenced before assignment
Besides, when I homogenized the model I was not notified that it was done successfully. I guess there might be a problem with the license file.
I would be grateful if you could address the problem.
Best regards.
Xin Liu @ on
Hi Reza,
If you just want to perform homogenization to get effective properties, you can use TexGen4SC at https://cdmhub.org/tools/texgen4sc , which integrates TexGen and SwiftComp. I think the Abaqus-SwiftComp developers may be able to answer the problem you met, but I am not sure if you can get the right results from the GUI since the material local orientations are defined in an unusual way in the output files from TexGen.
Reza Sourki @ on — Edited @ on
Hi Xin,
Thank you. Yes you are right.
I am looking forward to going deeper and further than performing homogenization so I need to import it into Abaqus-SwiftComp.
Have you tried importing a model from TexGen into Abaqus-SwiftComp?
-I tried to homogenize a simple cube modeled in Abaqus GUI, it worked. So, I believe there might be a problem with the model imported from TexGen.
Any ideas?
Xin Liu @ on
Hi Reza,
Usually, I directly use TexGen4SC to perform homogenization for textile composites. There are a lot of unnecessary definitions of nodal sets, steps and BCs in the inp file from TexGen for SC-Abaqus GUI, but if you want to use python script from TexGen website, then all the definitions are needed. I am not sure if this caused your problem. Another possible issue is the model information is stored in different files (at least in two files .inp and .ori), I am not sure if SC-Abaqus GUI can find all information. I am not an expert for SC-Abaqus GUI, may be the GUI developers can give you more accurate answers.
Reza Sourki @ on
Thank you Xin.
So, you suggest to carry out the homogenization in TexGen and then utilize the mechanical properties in Abaqus-SC GUI to carry out the analysis?
Wenbin Yu @ on
Reza, if you export TexGen into Abaqus file, then I will expect you to use Abaqus for the RVE analysis because that file is created for that specific purpose. Otherwise, you should export it to SwiftComp file, then you can either use TexGen4SC directly. If you want to use Abaqus-SwiftComp GUI (I don’t understand why), then you should use the SC file generated from TexGen4SC.
Please understand Abaqus-SwiftComp GUI is for you to perform homogenization and dehomogenization using a model created by Abaqus or a SC file created by other preprocessor such as TexGen or Gmsh. Abaqus-SwiftComp GUI does not perform macroscopic analysis. The properties out of homogenization should be used as inptus for regular Abaqus analysis.
Reza Sourki @ on
Thank you Professor Yu.
Please correct me if I am wrong.
My goal is to homogenize a unit cell which represents a SG of a woven fabric.
Then, use the properties to analyze a woven fabric (which I want to model as a 2D shell) under moment applied to one edge while the other edge is clamped.
Next, dehomogenize it to capture the local fields, stress and displacement fields, at any point of the structure.
All three steps can be done in Abaqus-SC.
Wenbin Yu @ on — Edited @ on
Reza Sourki @ on
Thank you Professor Yu.
Yes, exactly.
Reza Sourki @ on
Dear all,
I am trying to homogenize my model (unit cell imported from TexGen) in Abaqus-SC, but I assume homogenization is not carried out since I cannot import homogenized properties.
"Inputs echoed in file UnitCell_nSG3_2D_C3D6pbc.sc. ech!
Constitutive modeling for a 2D (plate/shell) model!
Homogenization will be carried out!
You are running SwiftComp Parallel Version!"
I would appreciate it if you could help me address the problem.