The method described is utilizing leaky surface waves. Due to the their nature, they are less than accurate. The requirement for surface conditions being just so to perform this test is inherent in the use of leaky rayleigh, lamb or sejwa waves. Cost effective and residual stress measurement are usually words that don't play well with each other for that reason. If your imparting the vibrations via a mechanical impactor, or a transducer, and receiving via mechanical/electrical means (except for acoustic transmission AE which is a different animal all together) out side of a lab for exacting finite measurement of stress levels, your never going to get it to work as it should in my opinion.
Same part, same stresses, same same across the board, simply changing the rms values of the surface will change the reaction of the surface waves and the accuracy of the results. Aerospace utilizes surface waves for examinations on a regular basis, but they typically have very strict surface finish requirements for that same reason.
To sum it up, the devil is in the surface conditioning. without exacting values for that, the test will only yield gross results, not accurate ones and then without a proper cal standard, not even that in my experience. In the real world, one welder to the next, the surface in the vicinity of the welds will vary, the welds themselve will vary, as will surface conditions due to coatings, handling, and other factors.
The theory behind your idea is sound, but I just can't see an accurate implementation due to extenuating circumstances.