Are you looking at non-linear analysis or are you expecting stresses close to YS? If you are trying to simplify and get something close then you could just model the parts w/o weld and assume the parts are bonded. If you need to get more 'real' you will have to approach the problem differently with non-pen contacts sets, etc. As a first pass I'd suggest KISS and see what areas light up.
model the welds as a separate part , do not use weld bead for FEA.
weld will be usually in triangle shape, so make right angled triangle with sides as weld thickness (sketch 1) and swept boss (sketch 2)