ASTM E2700 phased array weld testing: what the practice controls

What E2700 covers and where it stops

ASTM E2700 is a practice for contact ultrasonic examination of welds using phased array probes on ferritic material, principally carbon and low alloy steel plate and pipe. It covers sectorial scanning, where focal laws sweep a range of angles from a fixed probe position, and linear electronic raster scanning, where a fixed angle is stepped along the aperture. It addresses the scan plan, standardisation, sensitivity setting, scanning mechanics, evaluation and the content of the report, for both manual and encoded work.

It does not cover immersion testing, it is not a corrosion mapping practice, and it does not extend comfortably to coarse-grained austenitic stainless, nickel alloy or dissimilar metal welds. In those materials beam skewing, attenuation and grain scatter demand a technique qualified on representative material containing real flaws, usually with dual-matrix or low-frequency transmit-receive longitudinal probes. Specifying E2700 unmodified for a clad or overlay weld and expecting it to perform is a procurement error rather than an inspection failure.

Most importantly, it contains no acceptance criteria at all. E2700 tells you how to find, plot and record indications. Whether an indication is rejectable comes from the referencing code or the project specification, and the practice is explicit about that division of responsibility. Purchase orders that say 'PAUT per ASTM E2700' and stop there have specified a technique with no rule attached, and the argument about what to do with a 6 mm lack-of-fusion then happens on the pipe rack.

The scan plan is the real deliverable

Coverage in phased array is a geometry problem solved before the probe is placed on steel. The scan plan takes the weld preparation — bevel angle, root face, thickness, cap width, any counterbore or thickness transition — and demonstrates, with the beam set and index offsets to be used, that the fusion faces and the weld volume are swept by an adequate beam within the useful angular range of the sweep. Drawn properly it is an engineering deliverable, and a reviewer can find a coverage gap in five minutes that would otherwise be discovered during a fitness-for-service argument.

The failure is systematic and it always looks the same. A 40 to 70 degree sectorial scan from a single index offset covers a single-V weld in 12 mm plate comfortably. Put that identical setup on a 40 mm J-bevel and the upper fusion face lies outside the swept volume; the display is full of data, the operator sees a clean weld, and the region that mattered was never insonified. Two index offsets, a second scan set with different focal laws, or scanning from both sides is the fix, and all three cost minutes.

The other half of the scan plan is what happens at the extremes of the sweep. Beam energy falls and beam width grows at high refracted angles, so an indication detected at 68 degrees is not characterised with the same fidelity as one at 55. Procedures that treat the whole sweep as uniform overstate what the data can support. Getting the scan plan and the qualification block right before mobilisation is normally where ASNT Level III consulting earns its fee on a project.

Standardisation, sensitivity and the angle problem

Reference sensitivity is established on a block containing known reflectors — side-drilled holes at graduated depths, notches where the referencing code calls for them — after wedge delay and velocity have been calibrated. The step people skip is angle-corrected gain. In a sectorial scan the response from an identical reflector varies with angle because of wedge attenuation, refraction losses and beam spread, so the amplitude reference must be built across all angles used, not measured once at 60 degrees and applied to the whole sweep.

Time-corrected gain then compensates for attenuation with sound path, again at each angle. The consequence of omitting either is not subtle. A flaw on the upper fusion face, detected at a steep angle on a long sound path, can read ten decibels or more below its true amplitude, fall under the recording threshold and never appear in the report. Amplitude-based acceptance criteria are then applied to numbers that are simply wrong, and the weld passes on arithmetic rather than on condition.

Temperature is the quiet variable. ASME Section V requires the calibration block to sit within 25 °F (14 °C) of the examination surface, because both material velocity and wedge delay move with temperature. Standardising in an air-conditioned cabin and then scanning a line in direct sun breaks that limit routinely on site work, and the resulting depth and angle errors are systematic rather than random — every plot shifts in the same direction, which is exactly the kind of error that survives a sanity check.

Instrument, probe and encoder health

Phased array introduces failure modes conventional ultrasonics does not have. Individual elements die, and a probe with dead elements still produces a plausible-looking image with a distorted beam and a refracted angle that is not the one written in the procedure. ASTM E2491 gives the method for measuring element activity, and the rule most owners impose is that no adjacent elements may be inactive and the total inactive count must remain within a limit stated in the procedure. Run the check at defined intervals and after any impact, and keep the result with the probe serial number.

Wedges wear, and a worn wedge changes both the index point and the incident angle. On curved surfaces an unmatched or worn wedge leaves the leading elements poorly coupled, which quietly removes the steepest angles from the sweep. Verify the index point and the angles on a calibration block at the start of each shift and after any drop, and treat wedge radius as an essential variable whenever the scan is on pipe rather than plate.

For encoded scanning the encoder is a measurement instrument like any other. Verify it over a representative scan length — a metre is a reasonable check — and expect agreement within roughly one percent. Positional error puts a real flaw at the wrong weld station, which either sends a crew to grind sound metal or leaves the genuine defect in place while the repair is made somewhere else. Scanning speed matters for the same reason: exceed the acquisition rate and the file contains missing lines, so a limit on total data loss with no consecutive missed lines belongs in every encoded procedure.

Personnel: certification, demonstration and the analyst

A UT Level II certificate does not authorise phased array. The employer's written practice has to define phased-array-specific training, experience and examination, whatever the underlying scheme — SNT-TC-1A, ANSI/ASNT CP-189 or ISO 9712 — and many owners require a practical demonstration on a blind mock-up containing representative flaws before an operator is allowed near production welds.

This is a recurring audit finding on projects where phased array was substituted for radiography late in the schedule to avoid shutting down an area for a radiographic exposure. The equipment arrived, a procedure was written, and the operators carried general ultrasonic qualifications. Method-specific instruction to SNT-TC-1A is available through our NDT training and certification programmes, and on this technique the practical demonstration matters more than the classroom hours behind it.

The analyst is a separate skill from the scanner. Encoded data is usually evaluated from the file, often by somebody who did not perform the scan, and the ability to read a sectorial view alongside its corresponding B and C views, recognise root geometry and counterbore signals for what they are, and judge from the data whether coverage was actually achieved is what separates usable data from decorative data. Projects that budget for scanning and not for analysis get files nobody can defend.

Acceptance criteria come from the referencing code

For pressure equipment, ASME BPVC Section VIII Division 1 supplies amplitude-based criteria for ultrasonic examination, and Code Case 2235 provides flaw-size-based criteria where ultrasonic examination is used in lieu of radiography. Section V Article 4 and its mandatory appendices carry the phased array methodology that the construction code invokes; the construction code, not Article 4, decides what is rejectable and on what basis.

For pipeline girth welds, API 1104 gives workmanship criteria and an alternative acceptance approach founded on fracture mechanics and an engineering critical assessment, which shifts the whole basis from signal amplitude to flaw height and length. For structural work, AWS D1.1 requires an alternative ultrasonic technique such as phased array to be qualified and approved by the Engineer before it is used in place of the prescribed method.

Mixing the bases is the classic mistake. Amplitude criteria written for a fixed-angle shear wave examination, applied to a sectorial scan with no angle correction, produce sizing that cannot be defended in a dispute. If a project intends flaw-height-based acceptance, the procedure has to be qualified for sizing accuracy — normally against a block containing real or EDM-implanted flaws of known height — because a technique that detects reliably does not automatically size reliably, and the two capabilities are qualified separately.

Data, records and the audit trail

Encoded phased array produces a file, and the file is the record. It should be retrievable together with everything needed to interpret it: focal laws, probe and wedge serial numbers, calibration data, the scan plan, encoder settings, operator and analyst identity, date, weld identification and the evaluation. A PDF summary carrying three screenshots is not the record; it is a photograph of part of the record, and it cannot answer the question that matters two years later — was this region actually covered?

Owners increasingly require raw data at project close, and that requirement lands badly on contractors storing files on the instrument and a laptop in a site cabin. Holding scan files, procedures, personnel certifications and calibration records against the asset and the weld number is what an inspection data management system is for, and it is the difference between re-evaluation being possible and being theoretical.

Where an operator's data is contested — a rejected weld the contractor disputes, or a client questioning whether coverage was achieved — the file can be re-analysed independently. We carry that out as independent report and data validation, evaluating the scan against the code that was actually invoked in the contract rather than the one people remember invoking.

The misreadings that produce audit findings

Four recur across almost every phased array audit. Sensitivity established once at a single angle and applied across a sweep. Coverage claimed from a scan plan drawn for a different bevel or a different thickness. A calibration block outside the temperature window relative to the part. And a procedure whose essential variables — probe frequency, element count and pitch, wedge, angular range, scan index, gain — do not match what the instrument setup file actually contains, which is usually discovered by opening the setup rather than reading the paperwork.

The fifth, less often written up, is over-recording. An unqualified operator working at a 20% recording threshold with no geometry discrimination produces a report full of root geometry, counterbore and mode-converted signals, inside which a genuine lack of fusion looks exactly like everything else. Discrimination is a skill, and the procedure should state how geometric indications are identified and dispositioned rather than leaving it to judgement on the day.

None of these are exotic problems. All of them are visible in the procedure, the scan plan and the calibration record before a single weld is scanned, which is why procedure review before mobilisation remains the cheapest quality intervention available on any phased array scope — and the one most often skipped because the crew is already at the gate.

Sectorial scan coverage claims that outrun what the calibration actually proved

A phased array sectorial scan can display coverage across a wide angular range on the operator's screen, and it is common for a coverage map to be presented as evidence that the full weld volume was examined at every one of those angles with equal sensitivity. E2700's calibration requirements exist because that presentation frequently overstates what was actually demonstrated. TCG built and verified at only a handful of angles within the sweep, with the remaining angles interpolated by the instrument rather than independently calibrated, means sensitivity across the untested angles is assumed rather than proven.

This becomes consequential where flaw orientation matters most — a planar flaw oriented to reflect strongly at one angle within the sweep and poorly at the calibrated angles can be under-sized or missed entirely, while the coverage map for the scan shows full angular coverage and gives no visual indication that sensitivity was uneven across it. The scan record looks complete; the calibration record, read carefully, shows it was not verified as complete.

A procedure and report that comply with E2700 record TCG verification at each angle actually relied on for detection or sizing, not merely at a representative subset, and the examination report should make clear which angles were calibrated rather than presenting the sectorial coverage map alone as evidence of examination adequacy.

Request a consultation