What Each Nondestructive Method Can and Cannot See

Why this matters

Every nondestructive method has a direction it is blind in and a depth it cannot reach, and both are set by the physics of what the method responds to rather than by the quality of the equipment. So a clean report is never a statement that the joint has no flaws. It is a statement that no flaw was found of a kind, size and orientation that method could have detected at the sensitivity it demonstrated. Shops get hurt in one specific way: they buy the method they have used before, aim it at a flaw lying in the one plane it cannot resolve, and file the resulting clean report as evidence.

Before any examination: radiography makes the shot area a restricted area under 29 CFR 1910.1096, and nobody enters or works inside the posted boundary while a source is exposed; the source is licensed by the Nuclear Regulatory Commission or your Agreement State, not by the shop. Penetrant chemicals are commonly flammable in aerosol form, so use them away from ignition sources with the ventilation, gloves and eye protection each product's safety data sheet names. Magnetic particle prods can arc and burn both the part and the operator's hand, so use them only where an arc strike on the part is acceptable and keep contacts clean and tight. Where a surface has to be ground before it can be examined, that releases base metal and coating as respirable dust, an inhalation route needing capture at the tool or a respirator issued under a written program meeting 29 CFR 1910.134.

The gate, stated once

Radiography detects a flaw by how much material it removes along the beam path. Ultrasonics detects a flaw by the area it presents square to the sound beam. Those are two different questions about the same flaw, and they give opposite answers on the two commonest weld discontinuities.

A tight planar flaw lying nearly across the beam removes almost no material along the path, so it can sit under the detection limit of a radiograph while being a large, loud reflector to a correctly angled ultrasonic beam. A rounded volumetric flaw removes its full diameter along the path, so it is easy radiographically, while presenting an irregular scattering surface that returns a weaker and messier ultrasonic signal.

Every other capability in the table below follows from that sentence plus one more fact per method.

The methods side by side

The lack-of-fusion card owns what this table means for that one flaw specifically, and it is worth reading beside this one, because lack of fusion is where the gate above does the most damage.

Two things a method does not tell you, whatever it says

Sensitivity is demonstrated, not guaranteed. A radiograph's sensitivity is demonstrated by an image quality indicator, a machined shim and hole set of known dimensions placed in the shot. Reading a 2 percent thickness indicator means the technique resolved that indicator. A tight crack is a harder target than a drilled hole, so the demonstrated figure is a floor on technique quality, not a promise about flaws.

Coverage is a separate number from the verdict. An angle-beam scan from one accessible face covers less of a joint than a scan from both faces, and the report should say which. "No rejectable indications, 100 percent of length, one face" and "no rejectable indications, 100 percent of length, two faces" are different statements and shops routinely file them as the same one.

Choosing between the two volumetric methods when the code allows either

Where the governing document permits radiography or ultrasonics, four practical facts usually decide it before the flaw type does.

• Joint geometry. Fillets, T-joints and most corner joints are not radiographable in any useful way, because there is no clean two-sided path through the joint. That alone settles a large share of structural work.
• Thickness at the thin end. Ultrasonics has a near-surface dead zone, and on thin material that zone is a meaningful fraction of the section. Governing documents set a minimum thickness below which their ultrasonic provisions do not apply, so read that limit in the edition your contract invokes rather than assuming the method scales down.
• Material. Coarse-grained austenitic stainless weld metal and many nickel alloys scatter sound badly enough that a conventional angle-beam scan becomes unreliable. Radiography does not care about grain structure.
• Whether you need a record. A radiograph is an image somebody can look at in two years. A conventional ultrasonic scan is an operator's readings unless the scan is encoded and recorded, which changes what the file is worth in a dispute.

Worked example: one flaw, four methods

A 3/4 in plate butt splice, single-V groove with a 60 degree included angle, so each bevel face lies 30 degrees off vertical. The cap has been ground flush for radiography, so the radiographic thickness is the 0.75 in plate. Buried at mid-thickness on one bevel face is a sidewall lack of fusion, about 1 in long, with an opening on the order of 0.010 in.

Radiography, beam vertical through the plate. The gap is a plane inclined 30 degrees to the beam, so the beam's path through the gap is the opening divided by the sine of that angle: 0.010 divided by sin 30 degrees, which is 0.010 divided by 0.5, giving 0.020 in of missing material along the path. Against 0.75 in of plate that is 2.7 percent of the radiographic thickness. A technique demonstrating 2 percent sensitivity on an image quality indicator is nominally capable of that contrast on a machined target, so this flaw sits right at the limit rather than safely inside it. Call it marginal, and expect it to be missed as often as found.

Ultrasonics, angle beam from the plate surface. The bevel face is 30 degrees off vertical, so the normal to that face is 60 degrees off vertical. A 60 degree refracted shear wave, run from the correct side, arrives square to the flaw face and returns a specular reflection straight to the probe. That is the best case the method has. Swap to a 45 degree probe and the beam strikes the face 15 degrees off normal, and the returned amplitude drops. Run the same 60 degree probe from the opposite side and it now points at the other bevel face, not this one, which is why a scan plan names sides as well as angles.

Magnetic particle. The flaw is at mid-thickness in 0.75 in plate, well outside the shallow subsurface reach of the method. Blind by depth, before orientation even matters.

Penetrant. The flaw does not break the surface. Blind by construction, and a clean penetrant report here is correct and meaningless.

The score: one method sees it well and only from one side at one angle, one is marginal at 2.7 percent of thickness, and two are blind for structural reasons. A shop that owns a penetrant kit and a light has no route to this flaw at all.

What would flip the ranking. Replace the flaw with a slag inclusion 0.060 in thick lying in the same place. The radiographic path through it is the full 0.060 in, since a volumetric flaw does not have to be divided by any sine, which is 8 percent of the 0.75 in thickness and an easy find. The same inclusion is a ragged, irregular reflector that scatters sound in several directions, so the ultrasonic return is weaker and harder to size than the planar flaw's was. Same joint, same location, opposite method ranking, and the only thing that changed is the flaw's shape.

Check the example against the rules the siblings state

• The geometry is stated with its frame every time. Bevel faces are given as 30 degrees off vertical, refracted angles as 60 and 45 degrees off vertical, and the face normal as 60 degrees off vertical. No angle in this article is quoted from the horizontal, because mixing the two frames is how a scan plan gets written backwards.
• The arithmetic is re-derived. 0.010 divided by sin 30 degrees is 0.020 in; 0.020 divided by 0.75 is 2.67 percent, stated as 2.7 percent. 0.060 divided by 0.75 is 8.0 percent, stated as 8 percent.
• The rounding does not flatter the method. 2.67 percent against a demonstrated 2 percent sensitivity is called marginal rather than detectable, and the honest reading is that it will be missed some of the time.
• The sensitivity figure carries its basis. 2 percent is thickness sensitivity demonstrated on an image quality indicator, a machined target, not a crack.
• Direction check against the siblings. The lack-of-fusion card has radiography weak and correctly angled ultrasonics strong on sidewall lack of fusion. This example gives 2.7 percent radiographic contrast and a specular ultrasonic return, which is the same relationship with the numbers shown.

References

• AWS D1.1, structural welding code for steel, in the edition invoked by your project specification or the building code your authority having jurisdiction has adopted, for permitted examination methods and their acceptance criteria
• ASME Boiler and Pressure Vessel Code Section V, nondestructive examination, as invoked by the applicable construction code and adopted by your jurisdiction
• ASNT SNT-TC-1A or ANSI/ASNT CP-189, in the edition adopted by your employer's written practice, for technician qualification and interpretation level
• 29 CFR 1910.1096 for ionizing radiation and restricted areas, and 29 CFR 1910.134 for respiratory protection during surface preparation
• See related: What Lack of Fusion Is and Why It Hides; How to Inspect a Weld Without Destroying It; Why a Visual Inspection Finds Most of What Matters