You could use shrinkwrap to create a part without internal faces and then derive surfaces/faces into a new part.
But I wonder why your fea guys want surfaces for beam models. The math behind beam elements and shell elements is different and modelling a beam type structure using surface/shell elements can give incorrect results if the mesh is too course. (They should have a minimum of 3 shell elements across the web). It sounds to me that they are doing it the hard and time consuming way, and possibly incorrectly as well. But even if they want to use shell elements, the model might still be incorrect as the outer surfaces from your Inventor model will be in the wrong position for where the nodes should be. You should be giving them the mid planes. The significance of this will depend on thicknesses and aspect ratios.
I have come up with a solution (which doesn't completely resolve what I want but it'll do unless anyone knows of any other method) but I thought I'd post it on here for those who want a round-up on this issue..
The steps I have taken are:
1. I created my frame structure using Frame Generator, consisting of standard ISO I-Beam Sections, applying all necessary end treatments
2. I then highlighted all the beams and changed from the ISO section to my own custom section (which I published using iparts etc).. my custom section can be seen in the attachment 'I-Beam example'. - This off-set geometry allowed me to pick 3 surfaces from it (shown in red) which gives me the centre-planes of the web&flanges..
3. (The time-consuming part).. I then dervied (as a set of *orange* surfaces) each individual beam , saved as a new part, and deleted all the surfaces apart from the three mid-plane ones I wanted to keep.. so now i.e. representing my I-beam as 3 mid-section planes.
4. Then, in my frame assembly, right clicking and going component>replace, I replaced each beam with its derived surface identical.
This has resulted in me from having a frame produced in frame generator, to the same frame assembly, but the beams represented by 3 mid-plane surfaces, which represent the I-Beam, with all end treatments still intact.. (see attached 'screenshot1'. The reason I said it hasn't fully resolved the issue is that there is that the FEA guy has to extend the web surface to touch the two flange surfaces due to the section used (which cant be helped). However, its a process that will save my FEA team a lot of time with post-processing my frame generator model to use in third-party FEA software!
Hope this is useful to someone else!