Hi
In the school animation club I run we’ve got a number of Blender enthusiasts. One of the places we often hit issues is when they go to unwrap an object and hit problems with internal geometry; for example where they’ve merged or extruded and left faces trapped inside. My advice is to avoid internal geometry at all times, but am I right? Is it best practice to not have any, or are their circumstances where it’s acceptable, or even recommended?
Best wishes
andi
internal geo most of the time means non manifold topo
and this is bad for rendering
in general can cause problems of which faces to render ect.
also other tools might not work as intended !
there is a grid 3D that exist for some special cases
but depend what you do I guess !
happy bl
I can only speak personally. I can’t think of a situation where I’d actually want them, or where they’d be of use to me. Not in the way you’re describing it.
If you find something is causing problems where none should exist, then it’s not something you want in your models. And internal faces that run contrary to the overall topology are undesirable. The only circumstance where you want internal geometry is where those faces will at some point be visible. Where the normals are oriented to the viewer.
Internal geometry is common for 3d printing. It’s used as a method to reduce the amount of material used for a project. Essentially creating a hollow shell. But the normals are again oriented so they would face the viewer if they were to navigate the interior.
But in general use I have always avoided them, and taken care to do so. 
I am going to assume that you mean multiple shapes intersecting (like a cylinder coming out of another one for piping) and not unnecessary geometry like faces that can’t be seen or lose vertices. (Since those things always are a big nono)
For animations … if the part moves AND gets deformed due to the movement in the animation avoid intersecting parts at all cost.
For organic modeling i would personally also try and avoid it, since in many cases it will stick out like a sore thumb.
But sometimes it works … so if it feels like it fits then it probably is fine (altough that’s rare for me)
Then we have hardsurface modeling.
Honestly this is the place where lose but intersected geometry actually comes in very handy.
You already want (in many cases) very defined edges and by keeping a lot of the parts seperate you also gain more controll over each of them.
Plus a lot of otherwise difficult (or time intensive) shapes can be done a lot easier this way.
That said those “hidden” parts of the faces are going to screw with your uv mapping a bit and have to be considered when applying textures, alpha/normal maps and so on.
Basically you have to keep in mind, that just because a face isn’t completly visible in the end result, doesn’t mean that the uv map ignores it.
Edit:
In case that you mean geometry modeled in lets say a hollowed out part of a mesh … then only do it if realy necessary (as in that part is actually seen in the shot and can’t simply be modeled on the side and inserted into the animation via a cut to your “interior” model.
Not every scene of an animation has to be done with the same model.
By rendering such parts seperatly you also have the advantage of not having to worry about your interior geometry messing with the outside and you free up a lot of resources that can be used to improve the detail levels of the interior as well as the exterior shots.
And as a plus post processing those parts seperatly is much less of a hassle.
Basic rule is that if you can’t see it from any rendered angle , it doesn’t need to be part of the model.