The technique used the "twin-crystal probe" setting on the Ultrasonic Testing (UT) set, but plugging-in two separate leads to the connectors on the UT set, going to two independent identical single-crystal 60degree angle-beam probes. So they act as a transmitter-receiver pair (rather than a single probe working in pulse-echo mode). Defect detection would be from blocking the beam and reducing the transmitted signal (where in pulse-echo a defect causes a detected returning-beam). You would set-up the technique on a region of plate (defect-free) and produce a transmitted-beam signal amplitude say 80% of the screen height. Any defect in the beam-path would block at least part of the beam and reduce the transmitted-beam amplitude to less than the 80%.
The transmitted-beam technique's general limitation is you'd only know there are defect(s) there - but no indication of where they are.
On the other hand, in this specific situation, internal defects would be impossible to repair, so if defects were detected they would force remanufacture anyway. Nothing is being lost. The potential gain was if it could be demonstrated that there cannot be any flaws above a size which would make them unacceptable defects, the component could be saved - with big schedule-keeping benefits. Well worth having.
No Standard described and mandated the technique. The same general technique, but using 45degree angle-probes, is an obscure but mandated technique - but that angle would not be useful in enabling the required volume of the weldment to be inspected.
If the technique worked, we were going to have to use engineering judgment to insist on a "special" technique of our own application-specific specification.
We could visualise, and had already sketched, the ultrasonic test-block which would have to be made to demonstrate, justify and calibrate the technique. That could be readily made if the pilot trial suggested the technique would be useful.
In comparison, the single-probe pulse-echo technique is fast, accurate, efficient, informative and readily-performed.
We found the irregularities of the plate and weld-joint meant for every location you had to fix one probe position then hunt for any position and orientation of the other probe which would give the "80% reference level". So you are there with one hand fixed holding one probe, while the other hand does moves the other probe around - all covered in "goopy" (slimy and viscous) UT-beam coupling-fluid.
The concept of holding the two probes in a jig at a fixed separation facing each other - which was what the UT subcontractor also realised and where they provided of their own good-will a remarkable improvised but very serviceable device made from lumpy brass castings - could not be used.
Only if you could not find the "80% reference level" after an extensive search, positioning and twisting the probe (sweeping the beam), exhausting to any UT operator, could you say there seems to be defect in that location.
You would definitely not be using this technique if you had any other choice!
(R. Smith, 27Mar2017, 15Apr2017)