Hopefully you have modeled this as an assembly - that will make for fixing geometry for the FEA easier (IMHO). I would proceed with a globally-bonded model and chase down high stress points with solid-modeled welds with an appropriate cross section. Of course you will have to eliminate any interferences, small slivers and gaps prior to running this. The first places I see you might have problems are with the cope fitting at the top (handle/connection?) and from the round tube joint to the square tube.
look for the cosmos companion on weld analysis
stay away from it at first and follow james' recommendation to start with everything as bonded in a weldment/beam analysis