Things said about FEA that just aren't true 1
This is the start of a series that attempts to address some myths about FEA that are spread around the net by too many so-called FEA teachers who clearly just repeated them from their own teacher without verifying the truth by basic benchmarking.
Shell elements and thin-shell structures have two main myths repeated often by people who have clearly never either programmed or benchmarked Finite Elements.
Myth 1: Shell elements are more accurate for thin-shell structures than solid elements.
Reality
1: Firstly, any quadratic displacement element (with midside nodes) is better than any linear-displacement element (just corner nodes, also known as simplex elements). So do not use 3 or 4-node shell elements ever! In the late 1970's there was a raft of papers trying to create a new simplex shell element without inherent numerical errors and consequent spurious results but it was really just an academic exercise; all such problems disappeared by using quadratic shell elements.
2: Shell elements were developed as collapsed solid elements in the first place and they were derived not for accuracy but because the many simplifying assumptions meant they were quicker to solve. It may 'sound' correct to assume that thin-shell elements must be better for thin-shell structures but that is a fallacy.
3: Shell elements have a lot of extra differential geometry so suffer from several numerical errors unique to them; especially noticeable when modelling spheres by connected polynomial-shaped elements, where results are horrendous. For modelling spheres, using a 9-node Langrangian shell element is more sensible anyway.
4: Shell elements are good for large expanses of shell where you can likely calculate the stresses by hand anyway. But they are of little use at geometry changes, fillet welds and bolts; ie exactly where the high stresses are likely to be. Some tricks using beam elements and holes are often used to deal with these limitations; do not do that unless your methodology is thoroughly verified against solid elements or real-world tests!
Myth 2: You need 2, 3 or 4 solid elements through the thickness to model a thin shell.
Reality
Quadratic-displacement solid elements produce linear stresses (the derivative of displacements) hence if you know your through-thickness stress is simple linear bending, as in a thin shell structure, then a single quadratic solid element is sufficient through thickness, The only question that then arises is what is the maximum aspect ratio (length/thickness) . This can vary between 4/1 and 30/1 depending on the shell curvature but do your own benchmark test against an analytical solution of the unpierced vessel to verify. Close to the high stress regions, which are usually at changes of geometry you would need a smaller aspect ratio (ideally 1) but the mesher does this automatically for such locations. Note that consideration of the aspect ratio should also be done for shell elements but frequently isn't, leading to overly stiff models.
This myth seems endemic in the nuclear industry but is actually true only for simplex (linear displacement) elements which are presumably the only ones used by teachers of Finite Element analysis. Alas I have discovered that many engineers continue to use simplex elements, even though they were obsolete in the 1970's, because they allow huge models to run faster. But rather than use a badly-performing element it is far better to find a way to reduce the model size! Simplex elements are only barely useful for temperatures and displacements, and only potentially accurate at the centroid for stresses.
As a postscript, the Femdesigner Classic program has a unique shell element developed by me which has eight outputs; 2 membrane stresses in local x,y, in-plane shear stress, 2 bending stresses around x,y, twisting shear stress and 2 through-thickness shear stresses. Most shell elements combine these to reduce the total number of outputs to 5 so membrane stresses are only then found by averaging while in-plane and twisting shear are inseparable. Shell elements can still be very useful in certain circumstances but they are not vital nowadays thanks to the increase in computing capacity. If an older design used shell elements it was always due to computing limitations, not for accuracy.