Resolution size

Replies to This Discussion

Permalink Reply by Vadim Shapiro on August 25, 2010 at 11:59am

Excellent question. We have to be careful comparing resolution of mesh in classical FEA with the resolution in Scan&Solve. Remember that the whole point of Scan&Solve is that it works for ANY shape, not just simple shapes, and the elements in S&S do not conform to the shape. Imagine that you put your shape in a box, then mesh the box (not the shape), and pick the mesh elements that intersect the shape. If the box measures 10 cells across, you already have roughly 10x10x10 =1000 elements. This is oversimplifying quite a bit, but for some shapes you would want more than 10 elements across. Finally, the scan step of the solver, has to work fairly hard to adapt elements to the geometry at run time. It turns out that adapting to large cells is actually quite costly. Combined these considerations with the fact that S&S works for *any* solid geometry, our experiments suggest that 10,000 elements is a reasonable starting point for analysis.

We might be able to lower the 10000 to something smaller in the future, but so far our emphasis has been on improving the speed computation through advanced data structures and algorithms, while maintaining them as general as possible.

Could you post some of the examples you are looking at? I am sure everybody would be interest to know the results of your comparisons and conclusions. We could also run some tests that you cannot and see what happens.

• ▶ Reply

Permalink Reply by Hrvoje Petrovic on August 25, 2010 at 3:47pm

Thanks a lot for the answer, I am getting closer now to understand how S&S resolution works compared to a "usual" FE meshing. Here is a little more complex investigation which is in progress - a part of a structure we have designed which I will check against a FE software. I just have to find the time to do all of the export and meshing job to have a nice comparable mesh (here we go with the S&S advantages!). I thing the example will be good since we have huge stress peaks at certain points for which we are never sure how much they are caused because of meshing issues and how much because of unfavorable shaping of the part.

Attachments:

• Edge plate.jpg, 63 KB

• ▶ Reply

Permalink Reply by Vadim Shapiro on August 25, 2010 at 4:14pm

I hope that you will post your finding under "Solved by S&S" category in the forum. Also, if you would not mind posting this your jpeg file under the pictures, then everybody will be able to the example! Thank very much you for the interesting example and taking the time to do the comparison!

• ▶ Reply