Run87k wrote:
I'm not trying to rain on anybody's parade but it is very easy to get good looking but incorrect results from FEA. In the old days models were often verified with strain gauge results on test pieces or on the actual hardware but that seems to have gone away as it costs too much.
For comparing the effects of tubing sizes on stiffness, it isn't that hard to get it pretty close. Everything is a beam element and as long as you apply your constraints correctly you get something good enough for comparative purposes. Even if the final value is wrong, you can see that increasing that tube from 3/4" to 1" increased the stiffness by 5% - you're going in the right direction.
I don't disagree with the point you are trying to make (I really prefer hand calcs over FEA whenever possible) but what we are doing is pretty much the crudest FEA you can possibly do. I could probably create a system of equations for a space frame chassis and solve them by hand if I hated myself enough.
horizenjob wrote:
Additionally we are talking not about general FEA work here, but analyzing a space frame and even then in regards to the paper mentioned above and the program "Grape" that I use just doing beam and truss work which doesn't use a mesh at all.
Grape and LISA do a pretty good job of eliminating the mesh options from the user but by selecting your number of elements (not to be confused with the thing that goes between two nodes to represent a tube) you are essentially defining the mesh. In Grape choose File > Options > Element and select size from the drop down menu. LISA has a few other options under the Elements menu. Either way not touching it seems to work fine.
I haven't really played with the option much in either of these programs but in ANSYS the element size really didn't have much effect on the results for spaceframe chassises. I think the highest we typically went was 3. I'm guessing the tension/compression loading you mentioned really minimized its benefit.