Is there any literature saying something in this direction?
I found the extraction of effective permittivities of steel powers with microwaves 2.45Ghz DOI 10.1515/freq-2013-0083
Well, I know that frequency can change these properties and many other things like the particle size or oxide layers of course, etc.
But if I treat a powder layer like a semi-dense dielectric volume I would need to define some kind of bulk epsilon range for it - a barely plausible approximation?
Surly no way to approximate this even with rough values?
Electro magnetic testing or eddy current testing or using elkometer type instruments may be worth to think on your requirement. Precision of readings depends on the frequency, depth and permeability