DGV asked the question:
I wonder if the 3D Strad people can build a computer model that changes the graduation virtually and see how the violin responds. That will be cool.
I have done a bit of analysing with Finite element analysis, and have found that a real model would be beyond the abilities of my 1000 node ABAQUS student edition. My model is a little more detailed than Schelskie's stick and point model. But his model is equivalent to a shell model. (just consider the points as vertices of the shell triangel elements.) Nobody should be using it for anything other than rough idea of what the motion is for a given mode. I should ask Scleskie if any account was taken of the orhotropic material constants.. Or perhaps someone can contact him.
What I can do:
1:) make a model with no damping and poor graduations. The sales literature says that one can do anything that the large version does, except that it is limited to 1000 points or nodes. Unfortunately, I found that one thing does not carry over..... that is specifying a thickness at every node in a model using shell elements. That right away screws things up. One could make a single-layer solid model and map out the thicknesses at the corners of each element. These will be prisms, of course, and the number of nodes will be doubled. A decent model can be made with about 400 nodes (times 2 for the solid one with graduations)
2.) That is only for a top. But one can enter the various elastic constants for wood in an orthotropic orientation, that takes nine constants. These can be found in the US Agriculture release on wood properties.
3: With this, one can find the normal modes of a top. I also made a shell model with constant thickness top and back with bassbar, f-holes, and position for soundpost. This had strings and a rough bridge, sides and linings which don't take many nodes.
4: But normal modes are not the whole story. One would like damping constants.
5:) To drive the system with a sawtooth type wave, one needs to use a different section of the program which I have not learned. The output is graphs of excitaion vs time.
6:) The normal modes are given, but one would need the time-dependent version to find the relative response of different modes to find the actual vibration (less damping).
7:) There are many ways to represent damping. A simple way requires two parameters and may be adequate for a first approximation.
8:) I do not know if any "computer people) have done all of this. It would be a major undertaking.
There may BE no one who has done it.
9:) However, one can get good approximations for static loads. This might be handy to decide on longitudingal archings or transverse archings. (My model used curtate cycloid transverse archings, and these had a special program to be calculated from the height of the central arch and the width of the arching at that point. (I terminated these at the purfling.)
Anyone may add to this list or ask about anything. Please feel free to comment.