Writing a UT procedure that survives a chemical plant audit

Start from the referencing code, and from its edition

A UT procedure is not a self-contained document. It exists to satisfy a referencing code, and the referencing code decides the acceptance criteria, the personnel qualification basis, the calibration requirements and, in some cases, whether ultrasonics is permitted at all for the examination in question. In a chemical plant there are usually three codes in play at once. New and altered process piping answers to ASME B31.3. Pressure vessels answer to ASME Section VIII, Division 1, which takes its examination methods from Section V. In-service inspection of that same equipment answers to API 570 and API 510.

These do not agree with one another. B31.3 sets examination extent by fluid service category and states its own acceptance criteria for weld imperfections, and it requires examiners to be qualified under the employer's program. Section VIII takes acceptance from the construction code and the applicable appendix. API 570 concerns itself with condition monitoring locations, corrosion rates and remaining life, and it consumes thickness data rather than weld acceptance data. A procedure that opens with "in accordance with ASME Section V" and stops there has not told anybody which acceptance criteria apply.

Edition matters as much as code. Jurisdictions adopt specific editions, owner specifications freeze others, and the procedure has to cite the edition it satisfies. Writing to the newest published edition is not automatically correct; it is correct only if that is the edition in force for the work. The first question in any procedure review is which document, which edition, and who says so.

Essential variables, and the ranges nobody demonstrated

ASME Section V, Article 4 sets out the requirements of an ultrasonic examination procedure in Table T-421 and marks which items are essential variables. The concept is simple: change an essential variable and the procedure is no longer qualified until it has been demonstrated again. Change a nonessential variable and the procedure still has to be revised, but no new demonstration is needed. Most disputes about a procedure resolve into an argument about which side of that line a particular change sits on.

The failure pattern is almost always range inflation. A procedure is demonstrated on 1 in carbon steel plate at 2.25 MHz with a 45 degree wedge, and then written to cover 0.25 in to 6 in, in carbon and stainless steel, at 45, 60 and 70 degrees, with any probe between 1 and 5 MHz. It reads as flexibility. It is actually a claim about capability across a range that was never tested, and the extremes of that range are where the physics is least forgiving: near-surface resolution at the thin end, attenuation and beam spread at the thick end.

The discipline is to write the range you can defend and to accept that a chemical plant with genuine material diversity needs more than one procedure. Two well-qualified procedures covering distinct material and thickness families beat one heroic document covering everything, because the heroic document is refuted by a single question about the thin end of the range.

Damage mechanism first, technique second

Workmanship procedures are written to find fabrication flaws: porosity, slag, lack of fusion, incomplete penetration. In-service chemical plant examination is looking for something else entirely, and the mechanisms are specific enough that they should be named in the procedure. External chloride stress corrosion cracking of austenitic stainless under wet insulation. Internal chloride cracking in process service. Caustic cracking in alkaline service where the temperature crosses into the susceptible band. Amine cracking in gas treating. Polythionic acid cracking of sensitised stainless during shutdown, which AMPP/NACE SP0170 exists to prevent through neutralisation and purge practice. Hydrogen blistering and stepwise cracking in wet sour service. Microbiologically influenced corrosion in cooling water circuits. Erosion at elbows and tees in slurry and catalyst service. Dew point corrosion where a stream crosses its acid dew point.

Each of these produces a characteristic flaw with a characteristic orientation and location, and that dictates technique. Surface-breaking ID cracking is angle beam work from the outside surface, with a technique that can resolve near the far surface where the crack starts. Stepwise and hydrogen-induced cracking is mid-wall damage that a shear wave technique will scatter across and a straight beam mapping technique will resolve. General and localised wall loss is a gridded straight beam or encoded corrosion mapping job. Cracking at a weld toe on the inside of a jacketed reactor may not be reachable by ultrasonics at all.

API RP 571 is the bridge document, and a procedure that cites the mechanisms it is intended to detect immediately becomes more defensible than one that does not, because it has answered the question every good auditor asks: what is this examination for?

Materials that defeat a carbon steel procedure

Austenitic stainless weld metal is anisotropic and coarse grained. The beam refracts, skews and scatters, the noise floor rises, and a conventional 2.25 MHz single-crystal shear wave technique that works beautifully on carbon steel produces a display where a real flaw is indistinguishable from grain noise. The mitigations are known: lower frequency, dual-element transmit-receive longitudinal probes, and in demanding cases matrix array techniques with adaptive processing. None of them substitutes for demonstrating detection on representative weld metal.

Duplex and super duplex stainless behave differently again, and nickel alloys such as Alloy 20, the Hastelloys and Inconels each have their own velocity and attenuation. Titanium and zirconium equipment appears in acid service and brings its own constraints. Clad and weld-overlaid vessels add an interface that both reflects and attenuates, so an examination through cladding needs a technique demonstrated through the specific clad thickness and condition, and disbond examination is a separate technique with a separate acceptance basis.

The couplant issue follows directly from this material mix. Section V limits residual sulfur on nickel base alloys and residual halides on austenitic stainless and titanium, at the 250 parts per million level. The reason is contamination-driven cracking, not fussiness. In a plant with purity constraints there is a second reason: couplant residue on a surface that will contact product is a contamination event in its own right. The procedure must name the couplant, reference the certificate, and specify removal.

Calibration, blocks and transfer correction on real plant surfaces

The basic calibration block is where most chemical plant procedures quietly fail. Section V wants a block of the same material specification, product form and heat treatment condition as the component, with reflectors at depths bracketing the examination range, and with curvature addressed where component diameter makes a flat block unrepresentative. In practice, one carbon steel block travels with the crew and is used for everything, because the alloy blocks were never bought.

Transfer correction is the second gap. The block and the component almost never have the same surface condition or the same attenuation, and the difference is measured, not assumed: a pitch-catch measurement on the component compared with the same arrangement on the block, expressed in decibels and applied to the evaluation. This is essential whenever DGS is used, and it is good practice with DAC. It is also the single most reliable way to tell, from the paperwork alone, whether a contractor's ultrasonics is serious.

Instrument housekeeping completes the set. Section V requires screen height and amplitude control linearity checks at intervals not exceeding three months, or before first use thereafter, with records tied to the instrument serial number. Where examinations are performed outside the temperature band in which calibration was carried out, the calibration must be repeated at temperature or a correction demonstrated. Both are trivially auditable, and both are routinely missing on jobs where the technical work was otherwise sound.

What the qualification demonstration has to show

A demonstration is an experiment with a stated hypothesis: that this technique, on this material, in this thickness and surface condition, detects the flaw type of interest with adequate sensitivity, and where sizing is claimed, measures it within a stated tolerance. The specimen set has to reflect the plant, which means representative material and weld procedure, representative surface condition including any coating or mill scale that will be present, and flaws of the type and orientation the damage mechanism actually produces rather than a convenient row of side-drilled holes.

Where the result will feed an API 579 assessment, sizing is the point of the exercise, and the tolerance is a number that will propagate into someone's remaining life calculation. The demonstration should establish it against flaws whose true dimensions are known independently, and it should be repeated by a second qualified technician. If two technicians differ by more than the tolerance you intend to claim, the tolerance is wrong, not the second technician.

Coverage deserves the same treatment. A scan plan asserting 100 percent volumetric coverage of a nozzle weld is a geometric claim and it can be checked. Where geometry defeats access, the procedure should say which volume is not examined and why, so that the inspector receiving the report knows what he has and what he has not. Undeclared missing coverage is the most dangerous defect on this list, because it looks like a clean result.

Findings that recur when this procedure is audited

First, the procedure cites a superseded code edition, or cites Section V without naming the acceptance criteria document. Second, essential variable ranges exceed anything demonstrated, most often in thickness and material. Third, the calibration block does not match the component in material specification, product form or curvature, and no block certificate is on file. Fourth, DGS is specified with no transfer correction method described and no transfer correction records anywhere in the job files.

Fifth, a sizing tolerance is claimed with no demonstration behind it, usually in the same procedure that feeds fitness-for-service work. Sixth, the couplant is unnamed or uncertified against sulfur and halide limits on a plant full of stainless and nickel alloy. Seventh, instrument linearity records are missing or do not correspond to the serial number on the report. Eighth, the procedure is signed by a Level III who holds no method certification in ultrasonics, which is discovered by asking for one certificate and closes the discussion immediately.

None of these are exotic. All of them are closed by evidence rather than by argument, and all of them are cheaper to close before a client audit than during one. The point of a procedure review is to produce that list yourself, in writing, while there is still time to act on it.

Batch plants: writing for a four-hour window

Continuous plants give the inspection group a shutdown every few years with weeks of access and a full scaffolding budget. Batch plants give changeovers, and a changeover window is measured in hours, may be at elevated temperature, and frequently leaves the vessel jacketed, lined or full of internals. A procedure written for shutdown conditions is useless in that window, because the surface preparation it assumes cannot be done and the access it assumes does not exist.

The procedure has to be honest about this. It should state which locations are examinable in a changeover and which require a full shutdown, what surface preparation is achievable in the time, whether high temperature couplant and probes are required and to what temperature they are qualified, and what the fallback technique is when the primary access fails. It should also mark which readings are provisional, so a trend line is not built on data taken under conditions the procedure itself flags as degraded.

Jacketed reactors deserve their own paragraph in any chemical plant procedure. The jacket removes the outside surface as a scan surface for the vessel wall, and cutting inspection ports is a design and mechanical integrity decision rather than an NDT one. The realistic answer is often a combination of internal examination when the vessel is open, monitoring at the accessible nozzles and shell sections, and an explicit statement in the inspection plan that the jacketed portion is not being ultrasonically monitored. Writing that down is better engineering than a coverage claim nobody can meet.

Ownership, limits and how to start

Procedure development is supplied as a defined package: a review of the referencing codes and editions that actually apply, the procedures themselves with essential variables and ranges written to what will be demonstrated, the specification of calibration blocks and qualification specimens, the demonstration plan and its records, and the supporting revisions to the written practice where personnel qualification has to change to match. The procedures are yours, under your document control, signed by the Level III of record.

The limits are the same on this page as on the rest of the site. Atlantis does not act as the API 510, 570 or 653 authorized inspector for your equipment, does not sign in-service inspection records or set inspection intervals, and does not audit process safety management programs under 29 CFR 1910.119. Where a procedure question touches mechanical integrity, we supply the technical answer the program consumes.

The work is led by an ASNT NDT Level III certified in multiple methods who also holds API 653 authorized inspector certification, which means the procedure is written with an understanding of what the in-service code will do with the data it produces. To start with a review of your existing ultrasonic procedures against the codes and materials in your plant, request a consultation at info@atlantisndt.com. Scope follows the review and a quotation follows the scope.

Request a consultation