The first thing to consider is this - how do your model geometry and texture relate to each other? Given that you've modeled a fair portion of your assets, you should have a good idea of this from the get-go. I'd then establish this test: Find the FOV which is in FSO, and apply it to your 3D view in the modeling program of your choice. Create various points at reasonable distances from the model with which to determine when and where you need to apply the next LOD. When fine details can no longer be seen, it's time for the next LOD down - if the texture is good and is the same one used on the primary model, you'll never even notice the lower-detail LOD. As you very well know, the best LOD is one which is not noticed - you should not be able to tell when the lower-detail model or higher-detail model is swapped in for the other.
And of course, every model will have different LODs and distances at which those models are applied. Your best option is to go in and test every one individually... and that will be a lot of work. Otherwise, making size and detail comparisons between similar ships could cut out some additional time. In terms of speed, the best way to work with this is to test it in game, running with a similar approach as the static LOD distance test proposed above.
Last, there are multiple resolution sizes to consider. LODs on a 1024 x 768 monitor can be drawn in sooner because the resolution is lower than on a larger screen. I'm not sure what algorithm FS2 has built in to handle details in-game, but you should shoot for proper detail distances on a precieved monitor size which would be common for most users, where the assumption is also that they can run the program on full details. I personally would balance the LODs for monitors of a width between 1200 and 2000 pixels.
Note that this post was a bit late, and a few others popped up before I could complete it.