Same physics, different question
Both methods work the same way at the bottom. A piezoelectric crystal is excited by a voltage spike, it rings, and a pulse of high-frequency sound is pushed through a film of couplant into the steel. The sound travels, reflects off something, and comes back. The instrument times the round trip and multiplies by the velocity of sound in that material — about 5,920 m/s longitudinally in carbon steel — then halves it, because the sound made the journey twice.
That is where the similarity ends. What you point the probe at, what you expect the echo to come from, and what you do with the answer are entirely different jobs.
Ultrasonic thickness testing: how much is left
UTT points a straight beam at the far wall and measures the distance to it. The echo it wants is the back wall itself, so the answer is the remaining thickness at that exact spot, typically to ±0.1 mm.
It is the standard method for corrosion and erosion monitoring: process pipework, pressure vessels, storage tanks, silo shells, ship plate, structural hollow sections and anywhere a wall is quietly getting thinner from the inside where nobody can see it.
A few things matter more than most people expect:
- Probe choice. A dual-element (twin crystal) probe is used on corroded or pitted back walls, because a rough, irregular surface scatters the return and a single-element probe loses the echo. Single-element probes give better accuracy on clean material.
- Coating. Paint adds thickness and reads as steel. Echo-to-echo (multiple-echo) mode times the gap between two successive back-wall returns instead of the first one, which discards the coating and lets you measure through paint without grinding it off — a large practical advantage on a coated tank you do not want to breach.
- Surface preparation. Heavy scale, rust scabs or blistered coating have to come off. A reading taken on loose scale is a reading of the scale.
- Location, recorded. A thickness number is meaningless unless you can return to the same place. Readings are taken at condition monitoring locations (CMLs) on a marked grid, and the grid is part of the deliverable.
A single UTT survey tells you the wall thickness today. It cannot tell you whether that is a problem, because it does not know how fast the wall is going. That takes at least two surveys — which is a topic of its own, and the reason we wrote a separate article on setting inspection intervals from corrosion rate.
Ultrasonic flaw detection: what is inside
UT for flaw detection is not looking for the back wall. It is looking for everything that should not be there: cracks, lack of fusion, lack of penetration, slag inclusions, porosity and laminations.
A defect inside a weld is usually vertical or near-vertical, so a beam fired straight down often passes it without reflecting. Instead, an angle-beam probe launches a shear wave into the parent metal at a set angle — commonly 45°, 60° or 70° — so the beam travels diagonally, skips off the far wall and sweeps the weld volume from the side. The weld cap is never removed; the inspector scans the plate beside it and works out where in the weld the reflector sits from the beam path and the probe position.
The output is not a number. It is a defect's depth below the surface, its position along the weld, its length, and its echo amplitude compared with a reference reflector — a DAC curve or a DGS diagram built during calibration on a reference block. That comparison against an acceptance criterion in the applicable standard is what makes it a result rather than an observation. In Australia, weld ultrasonics is normally carried out to AS 2207, with acceptance to AS 1554 or whatever the fabrication code specifies.
It is also far more dependent on the operator. Thickness testing on clean plate is difficult to get badly wrong. Weld flaw detection is entirely a matter of interpretation — the difference between a geometric echo from the weld root and a real lack-of-fusion indication is a judgement call built on training and hours.
Where phased array fits
Phased array ultrasonic testing (PAUT) is flaw detection with the beam under electronic control. Instead of one crystal in one probe at one fixed angle, a PAUT probe holds an array of small elements — typically 16 to 128 — and the instrument pulses each one on a computed time delay. The wavefronts from all the elements combine, and by changing the delays you steer the resulting beam through a range of angles and focus it at a chosen depth, hundreds of times a second, without the probe moving at all.
What that buys you in practice:
- The whole weld volume in one pass. A sectorial scan sweeps, say, 40° to 70° continuously, so the inspection does not depend on the operator having chosen the right single angle for the defect that happens to be present.
- A recorded image, not a live call. Fitted with a position encoder, PAUT produces a scan tied to distance along the weld. The result is reviewable afterwards, by someone else, and comparable at the next inspection. Conventional manual UT leaves a written record of what the inspector saw at the time.
- Speed on long welds and thick sections, where manual UT needs multiple passes at multiple angles.
- An alternative to radiography. No radiation source means no exclusion zone, no permit, no night shift and no clearing the area of other trades — which on a live plant is often the whole reason it gets chosen. PAUT is usually paired with TOFD (time-of-flight diffraction) when accurate through-wall sizing of a crack is what matters.
The trade-offs are real: the equipment is expensive, the setup requires a written procedure and often qualification blocks representing the joint, and it needs an operator certified specifically in PAUT — a general UT Level 2 does not cover it. It is not the right answer for a handful of small welds where conventional UT will do the job in an hour.
Our own crews carry conventional UTT, visual, magnetic particle, dye penetrant and airborne ultrasound. Advanced methods — PAUT, TOFD, radiography and eddy current — we coordinate with partner inspection bodies, and we provide the rope access, rigging and supervision so the specialist does not need scaffold to reach the weld. In practice that means one mobilisation covers both the routine thickness grid and the phased array scan on the two welds that need it.
Which one do you actually need
- The wall is getting thinner and you want to know when to replace it — UTT, on a recorded grid, repeated.
- A weld has just been made and needs volumetric acceptance — UT (or PAUT/RT, depending on the code and the thickness).
- A crack has been found visually and you need to know how deep it goes — UT for sizing, TOFD or PAUT if the depth number has to be defensible.
- You suspect a crack but cannot see one — surface methods first (magnetic particle or dye penetrant, whichever suits the material), because they are faster and more sensitive to surface-breaking defects than ultrasonics is.
- A lamination in plate before you weld to it — straight-beam UT, which is closer to a thickness scan than to weld inspection.
Frequently asked
Can the same technician do both?
Not automatically. Certification is granted method by method under AINDT, ASNT or PCN schemes. Thickness measurement and weld flaw detection are separate qualifications, and PAUT is separate again. A Level 1 holder may acquire readings under supervision; a Level 2 sets up the technique and interprets the result.
Will thickness testing find a crack?
Only by accident. A straight beam aimed at the back wall will sometimes lose its echo over a large lamination or a badly corroded area, which is a clue, not a result. A vertical crack in a weld will usually give no indication at all on a thickness gauge.
Do you have to remove the paint?
For thickness testing on a sound coating, usually not — echo-to-echo mode measures the steel and ignores the paint. For flaw detection the surface has to transmit reliably, and heavy or blistered coating generally has to come off the scanning area beside the weld.
How thin can UTT measure?
Standard dual-element setups are comfortable from around 1 mm upward. Below that, single-element high-frequency probes and delay lines are needed, and the calibration matters far more.