Why is strain non-zero in this case?

FEAnalyst
Posts: 32
Joined: Sat Apr 03, 2021 5:28 pm
Answers: 1
x 9
x 22

Why is strain non-zero in this case?

Unread post by FEAnalyst »

Hi,

it's a rather simple issue but I'm not sure what actually happens here. Consider a rectangular plate meshed with 3D solid elements. The plate is placed between two rigid planes and I model it by simply constraining two opposite faces in all directions. The plate is subjected to uniform compressive force on all 4 side walls:
plate 1.JPG
Now what I don't understand is why strain in the direction normal to constrained faces is not zero:
plate 2.JPG
Analytical solution for this problem assumes zero strain in this direction. What's more, the resultant reaction force in this direction is different than it should be according to analytical solution.

Do you know what can be wrong here and what causes this non-zero strain in Z direction?

Thanks in advance for your help.
by jayar » Thu May 20, 2021 4:53 am
Perhaps you should try roller/slider constraints on both sides instead of fixed constraints. The fixed constraint would keep the edges of your solid from moving even though the sides want to move in. This should be especially noticeable at the corners since those points would be constrained in two directions.

Mind you - the roller/slider constraint in SWX doesn't always seem to produce perfect answers but they are very close usually.
Go to full post
User avatar
jayar
Posts: 121
Joined: Tue Mar 09, 2021 1:50 am
Answers: 2
Location: The Magic Kingdom (for now)
x 241
x 174
Contact:

Re: Why is strain non-zero in this case?

Unread post by jayar »

Perhaps you should try roller/slider constraints on both sides instead of fixed constraints. The fixed constraint would keep the edges of your solid from moving even though the sides want to move in. This should be especially noticeable at the corners since those points would be constrained in two directions.

Mind you - the roller/slider constraint in SWX doesn't always seem to produce perfect answers but they are very close usually.
FEAnalyst
Posts: 32
Joined: Sat Apr 03, 2021 5:28 pm
Answers: 1
x 9
x 22

Re: Why is strain non-zero in this case?

Unread post by FEAnalyst »

Thank you both for replies. With the help of forum members, I’ve found out that in order to correctly model this analytical problem using FEA (regardless of the software), different boundary conditions should be used than just fixed constraints. Slider/roller BCs on the side faces are a good idea. The only problem is how to constrain selected vertices to prevent rigid body motions but also obtain correct reaction force.

By the way, I use older version of SolidWorks (2019) so I won’t be able to open files created in newer releases and I would be glad for just the descriptions and screen-shots. I’m particularly interested in the scenarios with roller/slider BCs as those provide the correct reaction forces at the side faces.
User avatar
jayar
Posts: 121
Joined: Tue Mar 09, 2021 1:50 am
Answers: 2
Location: The Magic Kingdom (for now)
x 241
x 174
Contact:

Re: Why is strain non-zero in this case?

Unread post by jayar »

FEAnalyst wrote: Fri May 21, 2021 8:56 am Thank you both for replies. With the help of forum members, I’ve found out that in order to correctly model this analytical problem using FEA (regardless of the software), different boundary conditions should be used than just fixed constraints. Slider/roller BCs on the side faces are a good idea. The only problem is how to constrain selected vertices to prevent rigid body motions but also obtain correct reaction force.

By the way, I use older version of SolidWorks (2019) so I won’t be able to open files created in newer releases and I would be glad for just the descriptions and screen-shots. I’m particularly interested in the scenarios with roller/slider BCs as those provide the correct reaction forces at the side faces.
There is a trivially easy solution to this, two actually come to mind. The first is to fix just one vertex. This is my favorite as then I can check to see that the forces on this point (sometimes I use an edge if it would be a little more convenient in preventing rotation) tend to be very small. The other way it to try the inertial 'constraint' which is part of SWX.
FEAnalyst
Posts: 32
Joined: Sat Apr 03, 2021 5:28 pm
Answers: 1
x 9
x 22

Re: Why is strain non-zero in this case?

Unread post by FEAnalyst »

jayar wrote: Fri May 21, 2021 10:03 am There is a trivially easy solution to this, two actually come to mind. The first is to fix just one vertex. This is my favorite as then I can check to see that the forces on this point (sometimes I use an edge if it would be a little more convenient in preventing rotation) tend to be very small. The other way it to try the inertial 'constraint' which is part of SWX.
With one vertex fixed (and two slider constraints on side faces) the part tends to rotate about this vertex. The results are correct in this case (zero strains, accurate reaction values) but I would prefer to avoid this rigid body motion.
User avatar
jayar
Posts: 121
Joined: Tue Mar 09, 2021 1:50 am
Answers: 2
Location: The Magic Kingdom (for now)
x 241
x 174
Contact:

Re: Why is strain non-zero in this case?

Unread post by jayar »

FEAnalyst wrote: Fri May 21, 2021 10:22 am With one vertex fixed (and two slider constraints on side faces) the part tends to rotate about this vertex. The results are correct in this case (zero strains, accurate reaction values) but I would prefer to avoid this rigid body motion.
Using Inertial Relief might work just as well in that case.
FEAnalyst
Posts: 32
Joined: Sat Apr 03, 2021 5:28 pm
Answers: 1
x 9
x 22

Re: Why is strain non-zero in this case?

Unread post by FEAnalyst »

I usually don't use inertia relief due to concerns of potential impact on results. However, I do realize that in some cases (like floating or flying structures) inertia relief is a very useful feature.
Post Reply