I'm actually a bachelor student working on my final project-

We have a small wind turbine model built from a composite made of 50% e-glass fiber(material 1) and 50% epoxy(material 2). We divided our blade into 30 sections and for each we have a specific stitching pattern (0/+45/-45) for the leading and trailing edge, spar caps, and shear web. We want to get the material characterization of the composite in order to input them in ANSYS and FAST software that we use to run our calculations. However, we only have properties of the e-glass and epoxy.

We initially wanted to consider only one section to implement in SWIFTComp and we assigned nodes for our 3D model. I was wondering :

does the axis directions y1,y2, and y3 remain our proper choice and we are not constrained by SwiftComp ?

In user manual, we have a block mat_id isotropy ntemp. What does ntemp stands for ? is it the nodes number that are mode of the mat_id ? The micro3D.sc files seems to refer to it as such

in that same input file micro3D.sc , the last blocks refer to # mtype isotropy ntemp , for mtype=1 , we have a temperature density input that is not quite clear, what does it refer to ?

for mtype=2 in the latest block, we input 4 constants although isotropy=0, why do we have a line with constants 2600000 0.

Yes, the coordinate can remine your own proper choice. You are not constrained by SwiftComp. If you choose for SwiftComp is different from your problem coordinate system, you need to transfer SwiftComp outputs before input into your FEA code.

ntemp stands for number of sets for temperate related properties. If your properties are not temperature dependent, you can set ntemp=1.

temperature is a number for you to provide the material properties at a specific temperature, density is the density of the material.

2600000 0. implies the Young's modulus is 2600000 and Poisson's ratio is 0.

Thank you so much for your response. I had one last concern about the number of nodes allowed. I entered 300 nodes for my case, but it doesn't seem like SwiftComp can process it according to the ech file.

Also as every node is made of both material 1 and 2, I added # element material type & connectivity: element_no mtype node1 node2 , with same nodes for both materials, but the limit is 20 nodes. Is the choice of these 20 nodes irrelevant to the solution ?

I tried to run the software, but I get a blank .k file

Malak, for each element you can provide up to 20 nodes. There is no limit on mode number for the over model.
---- Emailed forum response from wenbinyu@purdue.edu

Actually, when I looked up the input eachoed in my file try.sc.ech , I noticed that not all nodes are included and that it stops running past 284 nodes when I input 300 nodes, and past 184 when I input 100 nodes (as you can see in the file attached below). I don't see how I can solve this issue. There are no error messages displayed in my output screen, only :

Inputs echoed in file try.sc.ech!

Constitutive modeling for a 3D model!

Homogenization will be carried out!

You are running SwiftComp Parallel Version!

Finished reading/processing model file!

Effective properties can be found in file try.sc.k! (which is empty whenever I check for output)

You need to learn some basics about finite element to deal with the input file directly. For example all 300 nodes must belong to the elements. Otherwise, you should use one of the GUI we have developed.
---- Emailed forum response from wenbinyu@purdue.edu

Thank you for your quick responses, I have uploaded the ANSYS SwiftComp GUI and I read the user manual. I'm trying to create a user-defined 3D model.I defined the materials, and generated the geometry by creating nodes that were read from a txt file (Screenshot of the geometry attached below). However, I can't find anything in the pdf or video tutorials about nodes.

I tried to create keypoints from nodes, but it doesn't really work for my case to have straight lines(trying to follow the example given in the user manual) as I have very close nodes, therefore I can't move on to the meshing (in the pdf, they use meshing from areas) and the homogenization.

I think you want to create mesh based on the nodes that were read from txt file, so it is a ANSYS APDL modeling problem. Maybe you can try this:

In the txt file, the I guess the codes are as follows: n, node, x, y, z (where node is the node number and x, y, z are coordinates). To create keypoints instead of nodes, just change 'n' to be 'k', i.e. k, node, x, y, z. This will create keypoints at the same place to replace nodes.

Then, you may want to create straight line based on those keypoints. The APDL command L,P1,P2 (where P1,P2 are two keypoints connecting those two lines) could make it.

Next, you may want to create area based on those straight lines. The APDL command AL,L1,L2,L3,L4,L5,L6,L7,L8,L9,L10 (where L1, L2, L3, . . . , L10 are List of lines defining area) could make it.

Finally, you could create mesh based on the areas for homogenization.

For more information about APDL command N, L, AL, you can refer to ANSYS APDL command reference: http://148.204.81.206/Ansys/150/ANSYS%20Mechanical%20APDL%20Command%20Reference.pdf.

Thank you Banghua Zhao, I followed your precious advices and I could create an area based on the straight lines. However, I'm stuck now with meshing options.

I don't know if it would be advised to follow a solid Quadractic 8 node 183 like they did in the user manual. Also for the manual size, I don't really what size I should take as I get warnings about the small lengths of my lines. As an alternative, I'm chosing basic smart size, but the meshing take ages and doesn't work due to element shape warning limits.

I think the size of each straight line is too small so that the smart size option gives you a very dense mesh (it may create at least one element along each line)

You could use lesize command to control the size of elements along each line. Just before meshing the area, use lesize,all,size (where size is what you set to control the size of each elements along each line, For example, if the total area is 100, I think a proper size would be sqrt(100)/10=1)

We are dealing with a pretty complex geometry and we haven't calculated its area, I tried to list it on ANSYS SwiftComp GUI, but the text file is not so clear an does not display the area and it shows an element size of 0. Also , when I tried to input sizes (for example I tried to set size =1) with a spacing ratio of 1 by defualt and then mesh, I have an error when running the solution , saying for example : quadrilateral element 19 has a zero or negative determinant of the Jacobian matrix at one of its sampling locations. Midside nodes, if any may be positionned poorly.

I wrote down the elements that have 0 or a negative determinant, but how can I eliminate them ? Also, is there a way to have the area calculated when I select it ?

You do not have to calculate the exact area. We could give an approximation of the area. We could also use the approximate length of the chord line, which may be more convenient. Say the approximated length of the chord line is 1. Then we could have lesize, all, 1/20.

I think you use alist to check the area. alist does not give the area but gives all the line that consist the area.

Also, before lesize,all,size, use allsel to select all the quantities.

I don't really get what you meant by allsel, is it when you Select > Everything ?. Other than that, I have tried to change the size to 1/20 and it still shows the same problem for the same specific quadratic element.

I don't know if I'm using the lesize correctly : Meshing > Size Controls > Manual Size > Lines > All Lines > SIZE= 1/20 and Spacing Ratio = 1.

I was wondering if I can eliminate the quadratic elements that represent a problem for my solution to run.

If you have APDL (ANSYS Parametric Design Lagrange) code (For example, the codes to create the areas: K, L, AL), that would be great. If you use ANSYS classical GUI, .log file would be okay.

I tried to create a fine mesh for your model but I still get the same error as yours: quadrilateral element has a zero or negative determinant of the Jacobian matrix. I checked the coordinate of the keyploints. I think the numerical values of the coordinate are too small so that they bring some numerical errors when calculating the Jaconbian matrix for each element in ANSYS. Maybe you could multiply each coordinate by 1000 (the unit will be mm in this case) to make the numerical values better for ANSYS.

Malak Yaaqoubi@ on — Edited @ onI'm actually a bachelor student working on my final project-

We have a small wind turbine model built from a composite made of 50% e-glass fiber(material 1) and 50% epoxy(material 2). We divided our blade into 30 sections and for each we have a specific stitching pattern (0/+45/-45) for the leading and trailing edge, spar caps, and shear web. We want to get the material characterization of the composite in order to input them in ANSYS and FAST software that we use to run our calculations. However, we only have properties of the e-glass and epoxy.

We initially wanted to consider only one section to implement in SWIFTComp and we assigned nodes for our 3D model. I was wondering :

micro3D.sc

988 B Click to download

Wenbin Yu@ onYes, the coordinate can remine your own proper choice. You are not constrained by SwiftComp. If you choose for SwiftComp is different from your problem coordinate system, you need to transfer SwiftComp outputs before input into your FEA code.

ntemp stands for number of sets for temperate related properties. If your properties are not temperature dependent, you can set ntemp=1.

temperature is a number for you to provide the material properties at a specific temperature, density is the density of the material.

2600000 0. implies the Young's modulus is 2600000 and Poisson's ratio is 0.

Malak Yaaqoubi@ onThank you so much for your response. I had one last concern about the number of nodes allowed. I entered 300 nodes for my case, but it doesn't seem like SwiftComp can process it according to the ech file.

Also as every node is made of both material 1 and 2, I added # element material type & connectivity: element_no mtype node1 node2 , with same nodes for both materials, but the limit is 20 nodes. Is the choice of these 20 nodes irrelevant to the solution ?

I tried to run the software, but I get a blank .k file

try.sc.ech

0 BClick to download

Wenbin Yu@ on — Edited @ onMalak Yaaqoubi@ onActually, when I looked up the input eachoed in my file try.sc.ech , I noticed that not all nodes are included and that it stops running past 284 nodes when I input 300 nodes, and past 184 when I input 100 nodes (as you can see in the file attached below). I don't see how I can solve this issue. There are no error messages displayed in my output screen, only :

Inputs echoed in file try.sc.ech!

Constitutive modeling for a 3D model!

Homogenization will be carried out!

You are running SwiftComp Parallel Version!

Finished reading/processing model file!

Effective properties can be found in file try.sc.k! (which is empty whenever I check for output)

try.sc

10 KBClick to download

Wenbin Yu@ on — Edited @ onMalak Yaaqoubi@ onDear Dr. Wenbin Yu,

Thank you for your quick responses, I have uploaded the ANSYS SwiftComp GUI and I read the user manual. I'm trying to create a user-defined 3D model.I defined the materials, and generated the geometry by creating nodes that were read from a txt file (Screenshot of the geometry attached below). However, I can't find anything in the pdf or video tutorials about nodes.

I tried to create keypoints from nodes, but it doesn't really work for my case to have straight lines(trying to follow the example given in the user manual) as I have very close nodes, therefore I can't move on to the meshing (in the pdf, they use meshing from areas) and the homogenization.

Thank you again

Banghua Zhao@ onHi Malak,

I think you want to create mesh based on the nodes that were read from txt file, so it is a ANSYS APDL modeling problem. Maybe you can try this:

In the txt file, the I guess the codes are as follows: n, node, x, y, z (where node is the node number and x, y, z are coordinates). To create keypoints instead of nodes, just change 'n' to be 'k', i.e. k, node, x, y, z. This will create keypoints at the same place to replace nodes.

Then, you may want to create straight line based on those keypoints. The APDL command L,P1,P2 (where P1,P2 are two keypoints connecting those two lines) could make it.

Next, you may want to create area based on those straight lines. The APDL command AL,L1,L2,L3,L4,L5,L6,L7,L8,L9,L10 (where L1, L2, L3, . . . , L10 are List of lines defining area) could make it.

Finally, you could create mesh based on the areas for homogenization.

For more information about APDL command N, L, AL, you can refer to ANSYS APDL command reference: http://148.204.81.206/Ansys/150/ANSYS%20Mechanical%20APDL%20Command%20Reference.pdf.

Malak Yaaqoubi@ on — Edited @ onThank you Banghua Zhao, I followed your precious advices and I could create an area based on the straight lines. However, I'm stuck now with meshing options.

I don't know if it would be advised to follow a solid Quadractic 8 node 183 like they did in the user manual. Also for the manual size, I don't really what size I should take as I get warnings about the small lengths of my lines. As an alternative, I'm chosing basic smart size, but the meshing take ages and doesn't work due to element shape warning limits.

Banghua Zhao@ onHi Malak,

183 element is advisable.

I think the size of each straight line is too small so that the smart size option gives you a very dense mesh (it may create at least one element along each line)

You could use lesize command to control the size of elements along each line. Just before meshing the area, use lesize,all,size (where size is what you set to control the size of each elements along each line, For example, if the total area is 100, I think a proper size would be sqrt(100)/10=1)

Malak Yaaqoubi@ onHi Banghua,

We are dealing with a pretty complex geometry and we haven't calculated its area, I tried to list it on ANSYS SwiftComp GUI, but the text file is not so clear an does not display the area and it shows an element size of 0. Also , when I tried to input sizes (for example I tried to set size =1) with a spacing ratio of 1 by defualt and then mesh, I have an error when running the solution , saying for example : quadrilateral element 19 has a zero or negative determinant of the Jacobian matrix at one of its sampling locations. Midside nodes, if any may be positionned poorly.

I wrote down the elements that have 0 or a negative determinant, but how can I eliminate them ? Also, is there a way to have the area calculated when I select it ?

Thank you so much for your help

Banghua Zhao@ onYou do not have to calculate the exact area. We could give an approximation of the area. We could also use the approximate length of the chord line, which may be more convenient. Say the approximated length of the chord line is 1. Then we could have lesize, all, 1/20.

I think you use alist to check the area. alist does not give the area but gives all the line that consist the area.

Also, before lesize,all,size, use allsel to select all the quantities.

Malak Yaaqoubi@ onI don't really get what you meant by allsel, is it when you Select > Everything ?. Other than that, I have tried to change the size to 1/20 and it still shows the same problem for the same specific quadratic element.

I don't know if I'm using the lesize correctly : Meshing > Size Controls > Manual Size > Lines > All Lines > SIZE= 1/20 and Spacing Ratio = 1.

I was wondering if I can eliminate the quadratic elements that represent a problem for my solution to run.

Banghua Zhao@ onYes, allsel means Select > Everything. I am not sure about the problem quadratic elements. Could you attach your APDL code?

Malak Yaaqoubi@ onWould you like me to attach .ech or .err file or .log ?

Banghua Zhao@ onIf you have APDL (ANSYS Parametric Design Lagrange) code (For example, the codes to create the areas: K, L, AL), that would be great. If you use ANSYS classical GUI, .log file would be okay.

Malak Yaaqoubi@ onI'm sorry I use the ANSYS SwiftComp GUI, so please find attached the .log file

works.log

4 KBClick to download

Banghua Zhao@ onSorry, I forgot to mention. Could you also provide your file, menust.tmp, to create the keypoints? Thanks!

Malak Yaaqoubi@ onI actually import nodes in a text file , then just create keypoints by picking all nodes

Nodes.txt

12 KBClick to download

Banghua Zhao@ onHi Malak,

I tried to create a fine mesh for your model but I still get the same error as yours: quadrilateral element has a zero or negative determinant of the Jacobian matrix. I checked the coordinate of the keyploints. I think the numerical values of the coordinate are too small so that they bring some numerical errors when calculating the Jaconbian matrix for each element in ANSYS. Maybe you could multiply each coordinate by 1000 (the unit will be mm in this case) to make the numerical values better for ANSYS.