The best evidence for grain flow on a really atomic scale comes from what is called texture analysis in X-ray or electron-beam crystallography (or related techniques): you get a deviation in the distribution of Bragg peaks due to the fact that you have a non uniform distribution over the orientation of the unit cells within the crystallites in the bulk material. You can fit this in a spherical harmonic basis and quite accurately work out the excess or defect of the distribution, typically quantified in units of 'multiple of a random distribution' or mrd, again either in crystallographic axes or traditionally in three orthonormal axes – parallel to the surface of the workpiece ("rolling direction"), axially transverse to it, and normal to it. The phrase to search for is 'pole plot'. They're rotationally symmetric and an inverse projection over all space, and so usually only a quarter of a hemisphere is shown.
A very good example of the affect of annealing tungsten wire is here [1] – note that (a) there is a very clear orientation dependence that some difficult geometric transformations will undoubtedly show means that they are aligned in the wire drawing dimension; and (b) after annealing at 1600 ºC for an hour the preference is slightly reduced but still about 15 sigma away from random...
I'm not saying grain flow doesn't exist. My claim is that I can't find any support for the idea that grain flow results in superior strength characteristics in the grain direction.
Ahh, I see! This is a common problem with things that are "known" to be true -- often people don't rigourously test them.
This paper [1] has some good data in it:
"The experiments in this study were developed to verify the influence of the grain-flow
orientation on fatigue life and its impact on the anisotropic properties of a mechanical
component. To this end, steel specimens were made, and their fiber was oriented by
machining and hot forging. Subsequently, they were subjected to flexo-rotational fatigue
tests in a piece of specific equipment to determine their fatigue life."
(...)
They then describe three parts: A, properly forged, B, improperly forged, and C, machined.
(...)
"The results showed that specimens of configuration A achieved a much longer fatigue life than configurations B and C, actually doubling it. The results indicated a similar fatigue life behavior between configurations B and C. It is important to emphasize that this similar behavior between these two configurations is valid for this case analyzed (...)"
A very good example of the affect of annealing tungsten wire is here [1] – note that (a) there is a very clear orientation dependence that some difficult geometric transformations will undoubtedly show means that they are aligned in the wire drawing dimension; and (b) after annealing at 1600 ºC for an hour the preference is slightly reduced but still about 15 sigma away from random...
[1] https://www.researchgate.net/figure/001-110-and-111-pole-fig...