It is possible, but it requires, that scattering on grains is measured. Maximum scattering should be observed, if the wavelength is ten times larger than grain size. Most standard ultrasonic flaw detectors are not sufficient for this measurement.
And, in addition, I think that the grain (crystal) orientation should be random. Only then you have macroscopic isotropy and thus it is negligible, wether the microscopic individual mechanical grain properties are isotropic or anisotropic.
If, in contrary, you have macroscopic anisotropy (texture) or dendritic grain growth (like in austenitic stainless castings or welds), then all is much more difficult, and grain sizing by UT might be impossible.
It's certainly possible.
There are some certain relations between the attenuation coefficient or sound velocity and the grain size. But as to different materials, there are different laws. and it seems that this subject has been being researched recent years.
Just as what Mr. Uli Mletzko said, if the material is macroscop anisotropy, it may become difficult to measure the grain size using UT. And we know, most of welds and casts are of columnar grain or dendritic grain, which means they are macroscop anisotropy.
I don't think so:
The largest metal grain sizes are typically of the orders of millimetre, and that's in slowly cooling castings. Most grains are of 10s of micrometre scales, specially in the relatively fast cooling environment of a weld. The grain boundaries are, usually a few atoms wide, ie nanometres. The smallest wavelengths of Ultrasound in metals are of the order of 100s of micrometres. It's impossible to distinguish separate objects of less than about 1/2 wavelength, so you will not be able to see that grain boundaries. Another reason is that they are, at best, going to be a bit like another shade of grey, with a slightly different refractive index.
X-ray tomography is able to pick out grains, but it has to be able to get all the way round a fairly small object - is that the scale you are thinking of?
Here's an example presentation from ESRF:
note the summary :
1008 reconstructed grains
2 days of processing (30 nodes)
approx. 3 micrometre accuracy
ESRF is NOT a portable measurement device: www.esrf.eu/AboutUs/AboutSynchrotron
You are writing, that: "It's impossible to distinguish separate objects of less than about 1/2 wavelength, so you will not be able to see that grain boundaries."
It is not correct. To be able to recognize the grain size you do not need to visualize them. It is fully enough to measure the scattering. And this occurs at wavelengths that are many times larger than grain size. It is possible to measure grain size with ultrasound and such measurements are known and used since years, but naturally not good in each possible case.
If this is applicable, the method is simple and quick - this can be the large advantage.
Of course it is possible to estimate the grain size by measurement of ultrasound attenuation in the reference blocks at different frequencies.
In Russia this method spreading widely, particularly it Railways (axis, wheels, rails materials)...
We have a special code for example (Ð? 153-34.1-17.404-00)