May be this is due to Solid elements having only three degree of freedom per node(i.e. traslations). But the shell elements having six degree of freedom per node (i.e. 3 rotational & 3 translations). Due to this you are getting different frequencies between the solid mesh and shell mesh while running the frequecy analysis.
As per my knowledge, first we need to decide which type of mesh we need to use. when the L/t ratio is more than 15 we need to use shell elements.
Kindly let me know if i mentioned any this wrong.
With the right mesh density you will get the same answers more or less regardless of element choice. I suggest that you do something that you know the answers to like a cantilever beam, fixed at one end. Do experiments till you figure out under what settings all of the analysis produce more or less the same results. They should all get the same lower modes and be reasonably close. Then you will know what it takes to get an answer with whatever elements and geometry you are considering. If this doesn't happen then there are some serious issues. I am pretty confident that it will though.
Have you figured out what is going on? Or at least the best approach to the most accurate solution?
I'm experiencing the same problem now. I have an assembly with sheet metal components.
By the guidelined, they should be treated as shells, but...
I have the option to treat the components as solids or shells. Using the same mesh settings, the two give vastly different frequency values.
Doing a mixed mesh, some solid and some shell, gives even different results. I'm using 09 sp4.0