Using a skeleton model as a placeholder for shell elements...

[xvis_2019]

We are having some problems with establishing a working FE-model of the following tank for a hydrostatic analysis (see first figure).

We are relatively new to Nastran, especially using shell elements. Due to plate bending etc (see figure 2 for a random illustration of typical problems with sheet metal), Nastran has a hard time making sense of it.

 

We have off course tried to make a simplified model, removing all unwanted details, even making a shrinkwrap with aggressive settings (removing fillets, pockets etc), but still Nastran will not make a working FE-model from it.

One solution we are pondering on for the moment, is to make a new, simplified model using a skeleton model (see figure 3) of the tank as input for Nastran.

And this is where we are a bit stuck. Would it be possible to use a skeleton model, and just tell Nastran which planes to use as shell geometry? That would simplify things a lot, no problem with seams and fillets etc. But we do not see quite how to go about doing the transition from the skeleton model to a Nastran model.

If we have to make a 'solid' shell from the skeleton, and convert all patches (surfaces where the stiffener are) to solid geometry before importing the model into Nastran, there is not much benefit to this.

Hopefully you understand what we are thinking of, basically the plane of the skeleton would be the the same as the 'geometry' of the shell elements. Is this conceivable?

[John_Holtz]

Hi @xvis_2019 . Welcome to Inventor Nastran.

What version of Inventor Nastran are you using? (In the Nastran environment, go to "Nastran Support > About").

I am not sure what you mean by a "skeleton" model, and I am not entirely sure what is shown in "tank_03.png". What you want to create in Inventor is a surface model. Instead of creating solids, you should create surfaces. The first branch of the Inventor model tree should be "Surface Bodies" instead of the usual "Solid Bodies".

Once you are in Nastran, you create Idealizations and set the geometry type to Shell. Then you choose the associated geometry (the faces) that you want to use for the idealization. With shell elements, the thickness and material properties of everything in the idealization is the same.

Let us know if you have any other questions.

[xvis_2019]

We are using the 2019-version at the moment (V13.2.0.168).

The last figure in the original thread shows the simplified model of the original tank. The tank is modeled as a solid, and the stiffeners are patches.  It is with this model we want to extract the surfaces and convert into shell elements.

Is there a way to extract all surfaces at once from a solid? For now, I exported  the model as a IGS-file and imported it back to a new ipt.

I am not sure if the term 'skeleton model' is correct, but a similar term has been used when making geometric placeholder for the frame generator.

[John_Holtz]

Hi @xvis_2019 . I do not know how to convert a solid to a surface in Inventor. It may depend on whether you want to use the outside surface (of the tank in your example, so that it would be in direct contact with the stiffener plates), or whether you want the mid-thickness of the solid converted to a surface.

@Anonymous or @Roelof.Feijen may know the answer. I am sure that their Inventor skills are far superior to my "skills".

[swalton]

The Inventor FEA (not Inventor Nastran) has a tool to find and convert thin parts to midplane surfaces.  Is there a similar tool in Inventor Nastran?  

Is is possible to use Inventor FEA to preprocess the thin parts to midplanes and then convert the FEA to a Nastran study for the hydrostatic loads?

[xvis_2019]

Nastran has these functions.

In our case the tank geometry was too complex in order to generate a working FE-model, using these functions. This is why we tried to create a FE-model directly from the surfaces of a simplified model.

[Roelof.Feijen · Accepted solution]

Hi @xvis_2019 ,

Sorry for the late response. We had a public holiday on Monday. Training on Tuesday. New laptop that needed to be set up. You know how it goes.

To convert a solid model into a surface model that you want to use in Inventor Nastran, you can use a number of features. If you use Delete Face on a solid without using the Heal option, a surface model is created. This is probably not the result you have in mind. Another option is Thicken / Offset. With the Offset you can offset the surface of the solid. However, this is very similar to the Midsurface feature within Inventor Nastran. Also not something I would recommend, but hey, you can.

The dilemma with FEA is always, will I use my production model (3D), of which I also make production drawings, for FEA calculations or will I make a separate FEA model for FEA analysis? I prefer the latter.

The basis for a Frame Generator model is in 99% of the cases a skeleton part with lines in 2D and 3D sketches.
(Complex) Sheet Metal parts can be made using a skeleton part based on surfaces and the Derive feature. So it is not surprising to make multiple parts to eventually make a production model (3D).

You probably already understand, I prefer an FEA model for FEA calculations. However, this does not mean that you cannot use the FEA model to create a production model (3D).

In this case I would make a surface model for the FEA calculations. Perhaps supplemented with lines in sketches for structural profiles. After you have completed the Nastran calculations, this model can be reused using the Derived feature. With this you transfer surfaces to individual parts that you reuse for making your production model / parts (3).
A big advantage of this method is that you can use the common origin technique, used in the car and shipbuilding industries, to create your assemblies without the need for assembly constrains, except for bolts, nuts, etc.

The problem with the term skeleton modeling is that it can be done in so many different ways. I therefore also prefer the term Top-Down design. You can use skeletons for Top-Down design, but you can also use multi-body workflows, or iLogic, or a combination of all these techniques.

I hope this helps.

[xvis_2019]

Many thanks for your reply. I think we have a similar approach to these problems. Separate the production model from a simpler analysis model is the way to go here.

After some testing I found that the 'thicken' function is one (of probably many) way of getting the surfaces from the solid model (of what I called the 'skeleton model'). Nastran has no problem with these surfaces.

My biggest gripe now, is whether it is possible to group model surfaces (tank, bottom stiffeners, side stiffeners etc). This what I am used to from previous FE-programs, so I hope this is a possibility in Nastran also. There is a lot of surfaces in this model, and they have to be selected over and over (mesh size, material/thickness, loads etc). It is time consuming and not very ideal in that you can easily miss a surface, or select a wrong surface. Is there a way of grouping items in Nastran?

I made a very basic analysis of the tank, and run into the following problem:

How do I identify where 'GRID 17909' is?

According to the manual some parameters (K6ROT, SHELLRNODE, etc) can be adjusted in case of this error, but it did not change anything. The K6ROT parameter, for instance, is greyed out.

[xvis_2019]

Added the FE-model if anyone would like to check it out.

[Roelof.Feijen · Accepted solution]

Hello @xvis_2019 ,

Turn Continuous meshing on in the mesh settings.

The nodes from faces and edges will match, so you don't need contacts.

If for some reason you need to use contacts with shell elements note that bonded will only connect the 3 translational dof and bonded offset will connect 6 dof.

[xvis_2019]

Many thanks! That did it!