Composites and Mechanica 1

[flyerfly]

I tried doing a verification model (simply supported beam) today with a composite sandwich and got very different results. Any thoughts?

Method 1: All solid elements were used for both the thick core material and the laminate. The material properties for the core is isotropic (foam) and the laminate is orthotropic. The problem with this is that meshing is poor for such a thin laminate compared to the core.

Method 2: Core was made with solid elements and laminate was done with shell elements (mid plane extraction from above model). The results between the two methods are drastically different (about 5.75x different!).

The deflection with the all solid elements was alot less then that of the shell+solid configuration. I am trying to figure out why and can't think of the reason so far. I would very much appreciate any insight into this.

Thanks!

 

[JohnAndrews]

I tried to create a similar model, but the deflection was pretty much the same in both.
Can you give dimensions and material properties and what version of Pro/E are you using.

 

[cpinz]

Hi Flyerfly
the problem of a 3d geometry with composites is not so trivial; your different results could be justified by the fact that solids and shell have different degrees of freeddom (dof) at their nodes; since the nodes must be common to both elememts in your structure, the element with lower dof governs the stiffenss of the structure. Mechanica solids have only translational dof so that deflection is more similar to a stair stepped (then more rigid..)curve rather than to a continuos curve (tipical of shells and beams).
More sophisticated software have 3d layered elements, but this is not the case of Mechanica, that works fine with shells for composites...also when you have sandwich structures.
I would suggest you to abbandon any idea to use Mechanica with 3d composites and concentrate your efforts on simulating your composites structures with shells....the most of people involved in the composites field( and I'm among them..) find this way efficient and trustworthy....

 

[flyerfly]

Over the last month or so I have found that using a solid core works well if I use the actual composite materials editor. This does not do the mid-plane extraction that mechanica does so well but requires the use of a surface. This surface is then assigned the lamination properties based on what you want. So far this has been working ok.

Years ago when I used nastran I could assign a moving coordinate system to follow the curvature of the part and thus get accurate results even on highly curved/compound surfaces. I think I just have to break the surfaces up into many regions so that the coordinate systems 3 direction remains substantially normal to the majority of the surface. If it does not then the error will go up for sure.

In your post above are you using shell elements for both the core and the lamination?

Thanks

 

[cpinz]

Yes I use shell elements...and until now I have not found reason to change to solid ...the structures that I analyze are typical of aeronautic field, that is, components where the ratio of lenght to thicknes respects the requirements necessary for shells. Besides I found very scarse literature that strongly raccomend the use of solids in the structural analysis of airframes or wings...and since I'm in the industry and not in the accademy I don't see any reason to try unuseful and not well documented works that justify their use. And in the composites field pushing a fem analisys too in deep is not worth the effort, as happens with metals..

You tell about nastran...I know it very well, and in particular the module of Patran for composites analisys, namely Laminate Modeler..I suggest you to try it

 

[flyerfly]

I would like to use Nastran and the Laminate Modeler but unfortunately our company does not have the money for that right now. I have not used Patran before...just looked at it.

When you use the shell elements in Mechanica are you using solids with mid-plane extraction to determine the shell or are you using only surfaces? When using the composite feature for shell properties in Mechanica it is not possible to use mid-plane extraction but I suppose I could do mid-plane for the core and not for the laminate. I'll have to do some tests...I think that our cores are too thick though.

Most of the panels that I am doing also respect the thickness/aspect ratios to support shell elements for the core as well but some do not. Some of the cores that I am using are pretty thick (up to 1.75 inches) and it is important to make sure that the core will not fail! I will try doing some verification runs to see how it well it works.

 

[cpinz]

I use only shell....but I never faced the problem of cores so thick. At this point transition between thin and thick cores becomes the problem, if you decide to mesh thin parts with shells and thick parts with solids; honestly i would insist with shells anyway testing a very simple case that can be succesively verified by hand ( e.g. simple sandwich beams oe plates loaded, as those that you find in the manuals of core producers...)