We frequently pull the inertia tensor out of Inventor (still using 2016) for assemblies as well as individual parts, and for parts which have overridden mass and parts which do not. We've noticed that occasionally we'll have a part with an intermittent calculation issue resulting in physically impossible inertia values (i.e. negative principle moments of inertia). When I check the global inertia values, they seem reasonable, but when I check principle inertia or inertia w.r.t. the CG, I see negative values where there shouldn't be (see attached image).
In the case of the part in the attached screenshot, I get invalid values when I check the part's inertia in the assembly (i.e. via context menu, so other assembly members are ignored), but opening the part by itself gives valid (and apparently correct) values.
When I calculate the expected CG by hand (assuming the global CG is correct), I get the same magnitudes that are given in the "w.r.t. CG" and principle inertia fields, but with the correct sign.
Any ideas?