cozzie55 Registered User
#1

Hey
I'm just doing some modelling using Ansys.
I was advised to apply a Hexahedral mesh to my models. Can anyone tell me why this type of mesh should be used?

I think it may have something to do with spare/box shape elements being easier and more accurate to predict strain with. But i want to be definite about this

Darren1o1 Registered User
#2

cozzie55 said:
Hey
I'm just doing some modelling using Ansys.
I was advised to apply a Hexahedral mesh to my models. Can anyone tell me why this type of mesh should be used?

I think it may have something to do with spare/box shape elements being easier and more accurate to predict strain with. But i want to be definite about this


Ansys help and theory guide is very helpful at explaining this.

CatFromHue Moderator
#3

I think triangles/tetrahedrals are less accurate than quads/hexs.

Another reason is that quads/hexs can distort more (better aspect ratio) than triangles/tetras.

Darren1o1 Registered User
#4

CatFromHue said:
I think triangles/tetrahedrals are less accurate than quads/hexs.

Another reason is that quads/hexs can distort more than triangles/tetras.


This is not true. The application of tri/tets is typically advised for certain geometries and can be more accurate. If one was simply less accurate than another, why would you not simply obsolete the less accurate element. It is specific to geometry and boundary conditions. For instance, implementing a mesh on a circle (say on a 2d crossflow) I would apply triangular mesh as I have have had better results in this case. Aspect ratio can be also controlled in the mesher for both types of element.

CatFromHue Moderator
#5

Darren1o1 said:
This is not true. The application of tri/tets is typically advised for certain geometries and can be more accurate. If one was simply less accurate than another, why would you not simply obsolete the less accurate element. It is specific to geometry and boundary conditions. For instance, implementing a mesh on a circle (say on a 2d crossflow) I would apply triangular mesh as I have have had better results in this case. Aspect ratio can be also controlled in the mesher for both types of element.


I presumed the OP was talking about FEA and was reading from Pg 203 of the book "Practical Stress Analysis with Finite Elements" which I think you're familiar with. Bizarrely enough in class today the lecturer brought up this very topic in specific to the Constant Strain Triangle Element (a linear three node element).

Darren1o1 Registered User
#6

CatFromHue said:
I presumed the OP was talking about FEA and was reading from Pg 203 of the book "Practical Stress Analysis with Finite Elements" which I think you're familiar with. Bizarrely enough in class today the lecturer brought up this very topic in specific to the Constant Strain Triangle Element (a linear three node element).


Sorry had heat transfer on the brain. I have found in practice that using a triangular mesh proximal to a complex shape, going, to quad gives the best result. Without knowing anything about the problem there is simply no way to argue for any case. I find the use of the help functions steers you well and gives good examples. If the OP would care to clarify the problem then I am sure we could be of help to them.

cozzie55 Registered User
#7

Darren1o1 said:
Sorry had heat transfer on the brain. I have found in practice that using a triangular mesh proximal to a complex shape, going, to quad gives the best result. Without knowing anything about the problem there is simply no way to argue for any case. I find the use of the help functions steers you well and gives good examples. If the OP would care to clarify the problem then I am sure we could be of help to them.


Rite will i'm trying to model simply supported 2m laminated timber beams, that have dowels inserted at angles through the laminates to joint them. So its a static structural problem. As it is a beam ie rectangular geometry, then i'm assuming a "square" mesh is more applicable than say a triangular one for most testing conditions.

The main reason i'm questioning the advise, is that largest mesh that the software applies is one with 48000 elements which takes the computer ages to solve(which is a pain when you have to run the solution a number of times changing parameters to get the rite solution). If i could use another mesh type the number of elements can be reduced ie an automatic method reduces it to 14000 elements.

Turbulent Bill Registered User
#8

cozzie55 said:
Rite will i'm trying to model simply supported 2m laminated timber beams, that have dowels inserted at angles through the laminates to joint them. So its a static structural problem. As it is a beam ie rectangular geometry, then i'm assuming a "square" mesh is more applicable than say a triangular one for most testing conditions.

The main reason i'm questioning the advise, is that largest mesh that the software applies is one with 48000 elements which takes the computer ages to solve(which is a pain when you have to run the solution a number of times changing parameters to get the rite solution). If i could use another mesh type the number of elements can be reduced ie an automatic method reduces it to 14000 elements.


Have you implemented any available symmetries or model reduction methods to reduce the number of elements required? You might only be interested in the beam's behaviour around the dowels (not the whole 2m length), so simulating the whole thing is a waste of time. You could run a really basic beam model just to ensure the boundary conditions etc. are right, then focus on an area.

In general, you need a really clear idea of what you want out of an analysis before starting one.

cozzie55 Registered User
#9

Turbulent Bill said:
Have you implemented any available symmetries or model reduction methods to reduce the number of elements required? You might only be interested in the beam's behaviour around the dowels (not the whole 2m length), so simulating the whole thing is a waste of time. You could run a really basic beam model just to ensure the boundary conditions etc. are right, then focus on an area.

In general, you need a really clear idea of what you want out of an analysis before starting one.


Yep i've already quartered the beam and am using zero displacement supports along the divided faces to reduce the size of my model and it still has 48000 nodes. This is due to the inclination angle of the dowels.

Darren1o1 Registered User
#10

cozzie55 said:
Yep i've already quartered the beam and am using zero displacement supports along the divided faces to reduce the size of my model and it still has 48000 nodes. This is due to the inclination angle of the dowels.


What about mapping your mesh focusing the mesh concentrations on the points of concern?

Darren1o1 Registered User
#11

Turbulent Bill said:
Have you implemented any available symmetries or model reduction methods to reduce the number of elements required? You might only be interested in the beam's behaviour around the dowels (not the whole 2m length), so simulating the whole thing is a waste of time. You could run a really basic beam model just to ensure the boundary conditions etc. are right, then focus on an area.

In general, you need a really clear idea of what you want out of an analysis before starting one.


+1

ZRelation Registered User
#12

If you throw up a picture of your part/mesh it would help in getting a better answer. As has been said, think carefully about how the mesh size can be reduced.

Which elements are better will depend a lot on your problem. It may even be a good idea to look at both tets and hex elements to see the difference in solution/ solution convergence (if even just to convince yourself of which is most suitable).

In my experience in general tet elements are usually discouraged except when necessitated by difficult to mesh geometries. Quadratic hex elements are typically recommended for elastic beams in bending.

Darren1o1 Registered User
#13

ZRelation said:
In my experience in general tet elements are usually discouraged except when necessitated by difficult to mesh geometries. Quadratic hex elements are typically recommended for elastic beams in bending.


Agreed.

#14

Could anyone suggest what might be the best Ansys element type to use for a thin sheet of metal (1mm thick) that's to undergo a large deformation?

CatFromHue Moderator
#15

Are you talking about for example a deep drawing process?

If so an axisymmetric shell element might be a good idea ;-)

Want to share your thoughts?

Login here to discuss!