Writing and Qualifying a UT Procedure for Cryogenic LNG Service

What the referencing code actually demands of a UT procedure

Start with the document the auditor will open first. ASME Section V, Article 4 sets out the requirements of an ultrasonic examination procedure and splits them into essential and nonessential variables. Essential variables include the material and configuration examined, thickness range, surface condition, couplant, search unit type and frequency, beam angles, calibration block and reference sensitivity, scanning technique and personnel qualification requirements. Change one and the procedure must be revised and requalified. Change a nonessential variable and the document is revised without requalification. Almost every argument on an LNG site about whether an existing procedure covers a new joint is really an argument about which column a change falls into.

The construction code above it decides what the examination is for. LNG process piping is normally built to ASME B31.3, where the examination requirements, the option to substitute ultrasonics for radiography with the owner agreement, and the ultrasonic acceptance criteria all have to be traced deliberately rather than assumed. Shop-fabricated pressure components come through ASME Section VIII, Division 1, whose ultrasonic-in-lieu-of-radiography route carries its own mandatory appendix requirements in Section V. Refrigerated containment sits with API 620 Annex Q and API 625, and the welding qualification underneath everything is ASME Section IX.

Where the technique is unusual or the material is difficult, ASME Section V, Article 14 provides the framework for qualifying an examination system as a whole rather than variable by variable. On LNG work that article is not an exotic option. It is frequently the only honest way to demonstrate that a technique chosen for nickel-based weld metal does what the procedure claims.

Cryogenic metallurgy is what really chooses the technique

Full containment tanks put 9% nickel steel to ASTM A553 Type I in the inner shell because it keeps fracture toughness at minus 162 C. The plate itself is fine-grained and ultrasonically well behaved. The weld metal is not. Those joints are commonly filled with nickel-based consumables in the alloy 625 and alloy 82 families, chosen for toughness and for tolerance of dilution, and the deposit is austenitic, coarse and strongly dendritic. A shear wave entering it loses amplitude, refracts away from the nominal angle along the columnar grains, and generates structural noise that sits at the same level as the reflectors you are hunting.

Cryogenic process piping in austenitic stainless behaves the same way for the same reason. The practical response is well established: drop the frequency to the 1.5 to 2.25 MHz region, use transmit-receive longitudinal or dual-matrix array probes so the transmitted and received paths do not both suffer, focus into the region of interest, and rely on diffraction rather than amplitude for through-wall sizing wherever the geometry allows TOFD. Amplitude-based sizing rules written for ferritic steel should not be carried into this weld metal without evidence.

The third material group is aluminium. Cold boxes and brazed aluminium plate-fin exchangers are largely outside the reach of volumetric ultrasonics, and a procedure that quietly implies coverage there will be caught. The honest procedure states the limits of the method by material group, names what covers the remainder, and does not sell a scan that physics will not deliver.

Cold surfaces: the measurement error runs the other way

Refinery inspectors learn that hot readings run thick. On an LNG plant the sign flips. Velocity in steel rises as it cools, transit time shortens, and an instrument holding the ambient calibration velocity reports less wall than exists, at roughly the same order of one percent per 100 F of difference. On a line running near minus 250 F and calibrated at 70 F, that is about three percent low. It is conservative in direction, which is exactly why it goes unchallenged for years, and it quietly generates false thinning trends that trigger investigations, repairs and schedule loss that the metal never justified.

The error compounds when the block is not the material. Velocity in 9% nickel steel and in austenitic stainless differs from carbon steel, so a survey performed with a carbon steel step wedge on stainless piping carries a material error on top of a temperature error, and the two do not necessarily cancel. The fix is unglamorous: calibrate on a block of the material being examined, verify velocity on a known thickness of that material, state the qualified surface temperature band in the procedure, and record the metal temperature with every reading so the correction can be reconstructed later.

Practical cold-work details belong in the procedure too. Standard couplants freeze; a low-temperature glycol-based couplant is needed and must meet contaminant limits. Probe wedges stiffen and lose coupling, and thermal shock cycling delaminates them. Frost and condensation form the moment insulation is opened, so the surface must be prepared and examined inside a defined time window. None of this is exotic, but a procedure that omits it produces data that cannot be defended when someone eventually asks how the readings were taken.

What the qualification demonstration has to prove

A qualification demonstration is not a formality and it is not a calibration check. It has to show that the technique, as written, detects the smallest flaw of concern at the least favourable position, in material that represents the production joint. Representative means the same material specification, the same filler and welding process, the same thickness, the same weld geometry including any backing or counterbore, and the same surface condition as the joints will actually have. A mock-up welded with a different consumable proves nothing about the technique you are trying to qualify.

The flaw set matters as much as the mock-up. Lack of fusion at the fusion line in nickel-based weld metal is the demanding target, not a side-drilled hole in the middle of a plate, because the fusion line is exactly where beam skew sends the energy somewhere else. The demonstration should include flaws whose position is not disclosed to the operator, a stated sizing tolerance evaluated against the true flaw dimensions, and a repeat by a second qualified operator to show the result belongs to the technique rather than to one person.

The output is a document, not a memory. Record the equipment and settings, the focal law set or probe and angle list, the scan plan with coverage plotted, the achieved detection and sizing results, the operator names and qualifications, who witnessed the demonstration, and the date and revision of the procedure it validates. On an LNG project the owner or EPC quality organisation will often insist on witnessing it, and refusing that request is a very expensive way to save a day.

Thick insulation, vapour barriers and where UT is the wrong tool

Cryogenic service is insulated heavily, and that insulation is a pressure-tight moisture barrier as much as a thermal one. Ultrasonics needs bare metal, so every in-service examination begins with a decision to breach the system. The procedure therefore has to own the whole sequence: where windows may be cut, what size, how the vapour barrier is removed and restored, how the surface is prepared without gouging the substrate, and who signs off the reinstatement. A UT procedure that stops at the probe and leaves the insulation to somebody else is only half a procedure.

The reason this matters is a genuine failure mode rather than a housekeeping preference. A poorly reinstated inspection window becomes a local thermal short. Moisture reaches the cold surface, condenses and freezes, and cycles between ice and liquid whenever the line warms. Corrosion under insulation on cold service concentrates precisely at those discontinuities and on lines that cycle in and out of cold service, not on lines that sit permanently at minus 160 C. The industry regularly finds its worst cold CUI at the ports cut to look for it.

That argues for sequencing methods rather than defaulting to ultrasonics. Screening approaches that see through insulation, such as pulsed eddy current or profile radiography, can narrow where the windows go, and ultrasonics then confirms and quantifies at a small number of justified locations. Writing that logic into the procedure, with the screening trigger levels that promote a location to a UT confirmation, is the difference between a targeted programme and hundreds of unnecessary breaches of a vapour barrier.

PHMSA, NFPA 59A and who ends up reading your procedure

A US LNG facility sits under 49 CFR Part 193, which governs siting, design, construction, operation and maintenance, and which incorporates NFPA 59A by reference for large parts of its technical content. Part 193 contains explicit construction-phase nondestructive testing requirements for welds, and PHMSA inspectors do ask to see the procedures behind them. That changes the audience for your document. It is no longer read only by a client quality engineer; it is read by a federal inspector who will compare what the procedure requires against what the weld records show was done.

Above the facility rule sit the process safety obligations. OSHA 29 CFR 1910.119 mechanical integrity requires inspection and testing on process equipment to follow recognised and generally accepted good engineering practice and to be performed by trained and qualified personnel, with the results documented. The ultrasonic procedure and the written practice behind the technicians are the artefacts that evidence the NDT portion of that requirement. Atlantis supplies that technical content; it is not a process safety management auditor and does not present itself as one.

The third reader is commercial. LNG projects are executed by international EPC contractors whose specifications frequently invoke ISO 9712 third-party personnel certification and European examination standards such as ISO 17640 or ISO 13588, alongside the ASME package the fabricator is working to. A procedure that satisfies ASME Section V and silently ignores the EPC specification will be rejected at document review, weeks after the technicians have been mobilised. Resolving that clash on paper, before mobilisation, is one of the cheapest interventions available on an LNG project.

Findings that recur when this procedure is audited

The same handful of findings appear across LNG fabrication and commissioning audits. The procedure was qualified on carbon steel and applied to nickel-filled 9% nickel welds. The calibration block does not match the material, curvature or heat treatment of the production joint. Transfer correction is required by the referenced standard but is neither performed nor recorded. Scanning gain and reference level are stated in the procedure but absent from the report, so the sensitivity used cannot be reconstructed. Personnel are certified in UT but hold no record of training or examination in the phased array technique they actually used.

The next cluster is about coverage and evidence. Scan plans claim full volumetric coverage without showing the beam coverage that supports the claim, particularly near the inner surface and at the weld root. Encoded data is collected but never archived in a form anybody can reopen two years later. Couplant certificates for halogen and sulphur content are missing on austenitic and nickel-alloy work. The metal temperature at examination is not recorded, so no correction can be applied retrospectively to a survey that later looks anomalous.

The last group is document control. Revisions are issued without a change record identifying which variable changed and whether it was essential. The Level III who approved the procedure is not certified in the method or is no longer with the company. Technique sheets in the field are a revision behind the controlled procedure. Each of these is trivially preventable at the desk and expensive to unwind once several hundred welds have been examined under the wrong document.

Request a consultation