I was wondering if the ANSYS Workbench SwiftComp module has been implemented to study plate (2D) and beam (1D) models. I have the following observations (or questions):
1. Regarding the homogenized properties for plates and beam models, I notice that the effective matrices change with the resolution of the mesh and some components do not show a convergence trend.
2. When importing the homogenized properties for the plate model I am not sure if the numerical values are consistent with the units of the model. I believe that SwiftComp expresses the homogenized properties using the SI system (m, kg, N or a combination of them). If that is the case, how do I ensure that these units are consistent with the dimensions and loads of the ANSYS structural model? Currently, when using the imported properties, I obtain really large deformations.
3. Has the MSG-beam structural model been implemented in Ansys? When I try to run the simulation with the effective properties from MSG, the simulation stops (shows an error)
I attach one figure that shows different effective matrices for different mesh resolutions of the same SG
I want to second your points. I spent 2 hours after work with my coworker and we found a lot of issues/questions with how the homogenization is being performed and what needs to be controlled.
For the plate model, the stiffness matrix had many odd values, such as 3 for the D11. Through some trial and error, manually manipulating the stiffness matrix value we saw that there is some issue with the order of magnitude (potentially unit driven) occurring. My coworker also helped me understand this Dmatrix can be influenced with the orientation of the laminate to the neutral axis, so he questioned how the homogenization was being performed and if how I modeled my SGs influenced the result.
We saw very large deformations with the plate model (2500mm) which were helped by manually manipulating the parts stiffness matrix, but ultimately calculated homogenized matrix was wrong.
We also were able to get the beam homogenization to run but only when we fixed displacement and rotation on one end and also fixed displacement on the other end of the beam. However, we saw HUGE deformation so this was clearly wrong as well.
Ultimately, without more clarity on how the GUI was programmed, including what units calculations are being performed in and understanding more about how the homogenization is being performed, I'm not sure how to troubleshoot the incorrect homogenization results I am getting.
I will begin to document the troubleshooting trials to help show my approaches to solving this problem, but I'm not confident I have the required skills to get the correct results with the information given now.
I have asked Mr. Banghua Zhao to help. Usually, to use a finite element code, you only need to be consistent with your units, then it does not matter what units you use. Of course, you need to understand the units of the quantities you are computing. For example, if you use mm for length, and N for force, then you need to input Young’s modulus using Force/Length^2, ie, N/mm^2, corresponding to MPa.
A: matrix is Force/length
B matrix is Force
D matrix is Force x length
---- Emailed forum response from wenbinyu@purdue.edu
In order to find more insights about the way homogenization is performed with the tool, I run the SG for the plate model assigning to the three bodies the properties of steel. It seems that for my model, SwiftComp outputs the homogenized properties in SI system (m, N, Pa). We have to be careful of using the correct units during the set up of the model and the dimension of the SG in Ansys. Therefore, I believe that the issue is not in the homogenized properties of the SG for the plate model but in the ANSY implementation of the structural plate analysis. When using the import option to perform the analysis with the homogenized matrices, I had large deformations. Also, there is a big chance I am missing something (or lots of things).
In the case of the beam model, I solved the problem by hand (following the notes from the lecture) and I got a tip displacement of 2.164 mm.
Homero,
I noticed this is my mesh convergence as well. Why would the number of elements change the sign (positive/negative) of parts of the stiffness matrix? This seems odd does it not?
In case you want to try, as long as you have meshed the SG, you can extract the information yourself to create the input file for SwiftComp according to the user manual of SwiftComp. The first several years, the students in this course did not have access to the GUI. They all create inputs themselves.
---- Emailed forum response from wenbinyu@purdue.edu
Hi Austin,
You are correct that the stiffness matrix is influenced by units. ANSYS Workbench introduced unit in the calculation but sometimes it is confusion. I suggest keep the same unit system across material properties and geometry. For example, if you use mm in solidworks to model geometry, then in workbench, use mm and MPa to be consistent.
For large deformation, usually it is because that transversely shear stiffness matrix is not given. If you have out of plane load, you need to supply transversely shear stiffness matrix. Otherwise, the stiffness in transverse direction is almost zero which causes the large deformation. For plate, use Reissner-Mindlin mode instead of Kirchhoff love plate model. For beam, use Timoshenko model instead of Euler-Bernoulli model. If the refined mode is not available, at least give non zero values to transversely shear stiffness matrix (c11, c12, c22).
Please let me know if you still have this problem. Thanks!
Banghua
在 2019年12月11日，上午9:08，Austin Graham @ cdmHUB - The Composites HUB <support@cdmhub.org> 写道：
---- Emailed forum response from zhao563@purdue.edu
For the different plate and beam analysis models, how do we change the setting from Kirchhoff to Reissner? Is this an option in the GUI or a hand calculation/manipulation we need to perform?
- The matrix of the figure was obtained assuming that everything is made of steel. I did that in order to understand the units that SwiftComp outputs. For my model, the homogenized properties are written in the SI system. By doing that at least I am sure that my problem is not in the homogenized properties but in the plate structural analysis. I still do not know how to solve it. I decided to go forward with the beam problem (solution by hand).
- Not sure what is the cause of not getting convergence.
The figure shows the SG and matrices I am using for the beam analysis.
I just double checked and made sure my coordinates were correct, my units were correct and have done a bit of changing the c11, c12, and c22 values to be non-zero. This unfortunately has not helped. I am still getting a 2872.3mm deflection with the effective properties shown. Note I have also changed the C11 and C22 values to 100 and the C12 value to 1.
It is an option in the GUI. In the homogenization settings, you can choose different models. The default is Kirchhoff for plate, Euler for beam. You can click it and it will be toggled to show other options.
Banghua
在 2019年12月11日，上午10:13，Austin Graham @ cdmHUB - The Composites HUB <support@cdmhub.org> 写道：
---- Emailed forum response from zhao563@purdue.edu
Banghua,
I reanalyzed my plate model with a fine mesh (0.5mm compared to 1mm) and achieved mesh convergence compared to (0.1mm); however, I have attached a picture showing that changing the order of magnitude of the c11 and c22 values directly changes the order of magnitude of the beam deflection. Increase c11/c22, and decrease the max tip deflection.
What value should be used for these "non-zero values"?
Homero,
Have you been able to use the extract result function in the swiftcomp GUI to dehomogenize the SG and extract the stresses for the plate model?
Homero Valladares @ on — Edited @ on
Hi,
I was wondering if the ANSYS Workbench SwiftComp module has been implemented to study plate (2D) and beam (1D) models. I have the following observations (or questions):
1. Regarding the homogenized properties for plates and beam models, I notice that the effective matrices change with the resolution of the mesh and some components do not show a convergence trend.
2. When importing the homogenized properties for the plate model I am not sure if the numerical values are consistent with the units of the model. I believe that SwiftComp expresses the homogenized properties using the SI system (m, kg, N or a combination of them). If that is the case, how do I ensure that these units are consistent with the dimensions and loads of the ANSYS structural model? Currently, when using the imported properties, I obtain really large deformations.
3. Has the MSG-beam structural model been implemented in Ansys? When I try to run the simulation with the effective properties from MSG, the simulation stops (shows an error)
I attach one figure that shows different effective matrices for different mesh resolutions of the same SG
Austin Graham @ on
Homero,
I want to second your points. I spent 2 hours after work with my coworker and we found a lot of issues/questions with how the homogenization is being performed and what needs to be controlled.
For the plate model, the stiffness matrix had many odd values, such as 3 for the D11. Through some trial and error, manually manipulating the stiffness matrix value we saw that there is some issue with the order of magnitude (potentially unit driven) occurring. My coworker also helped me understand this Dmatrix can be influenced with the orientation of the laminate to the neutral axis, so he questioned how the homogenization was being performed and if how I modeled my SGs influenced the result.
We saw very large deformations with the plate model (2500mm) which were helped by manually manipulating the parts stiffness matrix, but ultimately calculated homogenized matrix was wrong.
We also were able to get the beam homogenization to run but only when we fixed displacement and rotation on one end and also fixed displacement on the other end of the beam. However, we saw HUGE deformation so this was clearly wrong as well.
Ultimately, without more clarity on how the GUI was programmed, including what units calculations are being performed in and understanding more about how the homogenization is being performed, I'm not sure how to troubleshoot the incorrect homogenization results I am getting.
I will begin to document the troubleshooting trials to help show my approaches to solving this problem, but I'm not confident I have the required skills to get the correct results with the information given now.
Any help would be greatly appreciated.
Thank you.
Wenbin Yu @ on — Edited @ on
Homero Valladares @ on — Edited @ on
Hi,
In order to find more insights about the way homogenization is performed with the tool, I run the SG for the plate model assigning to the three bodies the properties of steel. It seems that for my model, SwiftComp outputs the homogenized properties in SI system (m, N, Pa). We have to be careful of using the correct units during the set up of the model and the dimension of the SG in Ansys. Therefore, I believe that the issue is not in the homogenized properties of the SG for the plate model but in the ANSY implementation of the structural plate analysis. When using the import option to perform the analysis with the homogenized matrices, I had large deformations. Also, there is a big chance I am missing something (or lots of things).
In the case of the beam model, I solved the problem by hand (following the notes from the lecture) and I got a tip displacement of 2.164 mm.
Austin Graham @ on
Homero,
I noticed this is my mesh convergence as well. Why would the number of elements change the sign (positive/negative) of parts of the stiffness matrix? This seems odd does it not?
Thanks.
Wenbin Yu @ on — Edited @ on
Banghua Zhao @ on — Edited @ on
Austin Graham @ on
Are the matrices shown what you got from the composite plate SG or what happened when you did the steel plate?
Austin Graham @ on
For the different plate and beam analysis models, how do we change the setting from Kirchhoff to Reissner? Is this an option in the GUI or a hand calculation/manipulation we need to perform?
Thank you.
Homero Valladares @ on — Edited @ on
Hi Austin,
- The matrix of the figure was obtained assuming that everything is made of steel. I did that in order to understand the units that SwiftComp outputs. For my model, the homogenized properties are written in the SI system. By doing that at least I am sure that my problem is not in the homogenized properties but in the plate structural analysis. I still do not know how to solve it. I decided to go forward with the beam problem (solution by hand).
- Not sure what is the cause of not getting convergence.
The figure shows the SG and matrices I am using for the beam analysis.
Austin Graham @ on
I just double checked and made sure my coordinates were correct, my units were correct and have done a bit of changing the c11, c12, and c22 values to be non-zero. This unfortunately has not helped. I am still getting a 2872.3mm deflection with the effective properties shown. Note I have also changed the C11 and C22 values to 100 and the C12 value to 1.
Banghua Zhao @ on — Edited @ on
Austin Graham @ on
It appears in my GUI it is a fixed option. There is no option to click. It seems locked.
Austin Graham @ on
Banghua,
I reanalyzed my plate model with a fine mesh (0.5mm compared to 1mm) and achieved mesh convergence compared to (0.1mm); however, I have attached a picture showing that changing the order of magnitude of the c11 and c22 values directly changes the order of magnitude of the beam deflection. Increase c11/c22, and decrease the max tip deflection.
What value should be used for these "non-zero values"?
Austin Graham @ on
Homero,
Have you been able to use the extract result function in the swiftcomp GUI to dehomogenize the SG and extract the stresses for the plate model?
Homero Valladares @ on
Hi Austin,
I have not tried dehomogenization given that I am getting really large deformation values for the plate with the homogenized properties.