Pradeepd's suggestion will give you a decent structural model of the cross, but it will not deal with the sif issue.
If the 'branches' are small in comparison to the header, I would contend that the relevant sif for a comparable tee appleid to each 'branch' would be appropriate. However if the 'branches' are anything like the header size, such that a bending load on one branch affects the other one, then you will need FE to get an sif.
You might visualise an FE plot of the loaded component ... if contour lines from one branch affects the other branch ... like affecting the 'back' of a normal three leg tee, then it's complex.