No, destructive testing isn't necessary.
There are two ways to determine the FN. One is to test the material with a magnetic meter, as indicated by Al and G Roberts. A well known brand of this meter is Ferritoscope.
The other is to use the WRC 1992 diagram, based on the chemical composition of the material. WRC 1992 goes from FN 0 to 100.
To be precise, FN is equal to ferrite percentage up to and including 8%. In other words, a FN of 8 corrisponds to 8%. Above this number, they vary, as G. Roberts has pointed out.
Giovanni S. Crisi
Sao Paulo - Brazil 

marswl,

to be honest, a longer time ago I wanted to know the same.

That was the reason that I have occupied myself with this interesting subject.

By the handwritten notes I have prepared that time please let me try to answer. Hopefully this hasn't been already done?

First of all, as been pointed out by the other forum fellow members, there appear to exist real severe problems in determining the exact ferrite percentage due to there exist a plenty of different methods in doing so.

Amongst others, in 1969 some valuable investigations were conducted (RATZ, G. A., GUNIA, R. B.: How Accurate are Methods for Measuring Ferrite? Metal Progr. 76, pp. 76-80, 1969). To keep it short, they have investigated how accurate the results were, which have been supplied or found out respectively, by 11 different companies and institutes, specialised in determining delta-ferrite contents. And the outcome proved it impossible to exactly determine the ferrite percentage. Neither by empirical formulae nor by the experimental methodologies. They could show that by up to a total content of 10% delta ferrite the scatter band amounted to +/- 3%. For total contents of 10… 24% whereas, the variation amounted to +/- 6%.

Another investigation with respect to approve or falsify these results was carried out by the IIW between 1971 and 1973 (Final Report on the Cooperative Testing Programme 1971 – 1973 on the Measurement of Ferrite in Austenitic Stainless Weld Metal , Welding in the World 12, pp. 17 – 26, 1974). Providing an approximate value of 3.5%, the measurements laid between 2.0% and 6.5%. Providing an approximate value of 7.5%, the measured results were stated between 5.5% and 12.5%. The final benchmark was set with an approximate value of 15.5%, and the scatter band laid between 11.9% and 29.0%(!). Other investigations have confirmed these partial massive variation and it could be shown that the higher the delta ferrite content, the greater the scattering. So far, and of course, this confirms what the other fellows have stated herein before.

Coming back to the question where the Ferrite-Numbers do come from or why they are dimensionless. Well, as so often our most appreciated fellow member Al (803056) has already shared the solution with us. Therefore let me quote what he wrote at that time:

“Ferrite numbers are measured using a "calibrated" gage that measure the amount of magnetic force attracting a magnet toward a sample.”

This method has been suggested by W.T. DeLong in 1974. And as far as my notes are correct, it came directly from the USA by have been published in (DeLong, W. T.: Ferrite in Austenitic Stainless Steel Weld Metal , AWS Welding Journal 53, Research Supplements, pp. 273 – 280, 1974). It has, to the best of my knowledge, even been taken over by the IIW and been published as  the ‘Standard’ method to specify the delta ferrite in a different way than before (percentage determination). DeLong’s methodology was based on what 803056 has so excellently described. A defined or “calibrated” permanent magnet is used to attract an austenitic weld metal sample. Simply put, the higher the delta ferrite content of this sample is, the higher the tear-off force to separate the sample from the magnet. This is a real improved way, since the force is not affected by any other or additional phases, eventually also originated by delta ferrite but even being non ferromagnetic. 

In other words, the tear-off force is an indirect value for the weld metal’s delta ferrite content. And now coming slowly to the fact that the FN are ‘dimensionless'. The tear-off force gauging is based on internationally agreed standards. I must apologise, but in Germany it’s called ‘Primärer Eichstandard’ and it’s hard for me to translate this into a proper English term. Anyway, the gauges used for calibration are specimen made of unalloyed steel, clad by copper layers of different thicknesses. Copper however, is non ferromagnetic. In other words. The thicker the copper layer, the lower the tear-off force. And here we go. A particular copper layer thickness – and here we are again with ‘GRoberts’ statement saying: “FN is an arbitrary scale that measures the magnetic response of the material…” – is now allocated to a particular FN.

According to my notes e.g. a copper layer thickness of 1.778 mm (0.07”) is allocated to the Ferrite Number ‘FN 3’. A layer thickness of 1.194 mm (~ 0,047”) is allocated to FN 5 and finally (at least according to my notes) a layer thickness of 0.610 mm (~ 0,024”) corresponds to the Ferrite Number ‘FN 10’.

Ah, I almost forgot. Not knowing if this is true, but at least it’s quite well imaginable. As it must be accurate, certainly it must, the gauges shall be prepared in the United States of America with the National Bureau of Standards.

Hope this helps.

Stephan