determining residual strain

[eyeguy]

Hello,

This is more of a general theory question rather than a technical one. I have full 3D data (to be specific, MRI scans) of a piece of tissue both in situ and after it has been cut away. Cutting away the tissue relieves any residual strain left in it, causing a deformation of the tissue.

What I am wondering is if anyone is familiar with a method (or a publication describing one), to reverse-engineer the residual strains in the tissue sample based on its shape before and after (or the deformation between). I have found a few papers, but they all deal with analytical solutions or using a single high order element. Has anyone ever done this using more traditional FEA--preferably ANSYS--before? The reason I ask is that it seems ridiculous to throw away a highly refined, anatomically-correct mesh and instead use a single element.

Thanks.

 

[cbrn]

Hi,
conceptually, with FEA it could go like that:
- create meshes in the two samples which are morphologically identical (same number of nodes, same number of elements, same connectivity between elements, ...); ideally let's assume that the nodes are even numbered the exact same way.
- get the coordinates of the nodes in the first sample, and in the second sample, then subtract node-by-node in order to have the displacements of every node. Let's assume you do the subtraction as "in-situ - removed"
- apply the "D" boundary conditions on every node of the "in-situ" model
- this should stress / strain the model; the strain is the relative strain between the two states.

I think the really difficult thing is to 100% control the meshes in order for them to be isomorphic...

Regards
-

 

[eyeguy]

I was thinking something along similar lines, but wondered why no one else seemed to do it. Now I know. Exactly numbering things will definitely be difficult.

Perhaps what I should do is start with three meshes: the two meshes that I already have from the data, and a third idealized mesh for which I know the nodes numbers etc. I could then somehow warp that third mesh to match the shapes of each of the other two meshes.

Does anyone know of a way to deform a mesh to match the shape of another object?

Thanks very much for your help.

 

[cbrn]

Hi,
aargh!!! In Ansys, I really don't know how to do that except, perhaps, with the instruments of Advanced Meshing in WB v.11 suite (I don't have A.M., so I don't know).
I seem to know there are some mesh-morphers on the market, probably you could restrict your search by first contacting your Ansys reseller and asking him for an advice.
Another way is to build an APDL script with which you "manually" build the mesh after you have recognized the locations of the desired nodes with a CAD:
CAD model of your meshes -> retrieve coordinates of the intersections of the curves -> create nodes at these locations, by assigning coherent numbers -> create elements through these nodes, paying attention to the order of declaration, and assigning coherent elem numbers.

Of course, before attempting anything like that, you must be sure that the number of "nodes" in your samples are both the exact same.
If not, then:
build a solid from the points clouds of both samples -> slice each one with the same number of isoparametric "u-v" curves (you need a very capable CAD in order to do that, UG NX or Catia are two examples...) -> retrieve coordinates -> etc as before.
Re-thinking about that, I think it will be your most probable way of proceeding.

I fear it's really not easy in both cases... You could also look for external services who perform reverse-engineering: they should have adequate software tools in order to do the first part of your processing (i.e. creating two coherent "node files" you will then be able to use for the calculation of displacements).

Regards

 

[eyeguy]

Hi cbrn,

Thanks for another very informative post.

It's unfortunate that you've confirmed what I suspected--that doing this won't be at all easy.

Unfortunately, external services likely aren't an option to me as I work with an academic research group affiliated with a hospital rather than at an engineering firm. It's unlikely I would be able to secure funding for an external firm to do work for us.

With that in mind, it looks like I may be stuck with the very slow process of building up a model on its own. I actually do have access to the mesh morpher feature in ANSYS, but it does not really seem to offer modifications beyond simple offsets etc. Given that I have complex geometry coming from a biological sample, it doesn't look like that will be good enough.

You don't know the names of any of the mesh morphing techniques do you? I am wondering if I could at least read about some of this stuff in academic literature.

 

[cbrn]

Hi,
no, sorry, this matter is completely outside my competence. Even if you can't hire an external service, you could try to look for info about "reverse engineering", "point-cloud interpolation", "free-form surface modeling from 3D-data" and things like that. I'd definitively choose the way to CAD-operate on the models first, and build an APDL afterwards.

Regards

 

[meshparts]

Hello,

Ansys has a very efficient result interpolation function, see *MOPER,,,MAP. With this function, one should be able to implement just what eyegui wants.

Regards
Alex