Writing and qualifying a UT procedure that survives a mill

The referencing code decides the procedure before you write a word

Steel and primary metals plants are code-plural in a way refineries are not. A single site can carry ASME Section I power boilers on waste heat recovery, B31.1 steam piping, B31.3 process piping for oxygen, nitrogen and gas cleaning, Section VIII vessels on utilities and air separation, AWS D1.1 structural weldments in the melt shop and casting bays, and a large population of non-code assets — ducting, cyclones, hoppers, hot blast mains, stove shells — that are inspected because they fail expensively, not because a code compels it. The procedure has to declare which of these it serves, because the acceptance criteria, the calibration requirements and the personnel requirements differ.

The non-code population is where procedure writing gets interesting and where most plants have nothing written down. There is no code acceptance criterion for a dust-eroded duct elbow; there is a remaining-wall criterion the plant's own engineering has to set, and a procedure whose job is to measure against it defensibly. Writing that procedure means importing the discipline of Section V — controlled variables, defined calibration, documented sensitivity — into work no code inspector will ever review, precisely because nobody else is going to catch the error.

The first question in every engagement is therefore blunt: for each asset class, what document are we writing to, and who accepts the result? A procedure that cannot answer that in its scope statement is a technical essay, not a controlled document.

Essential variables and the discipline of writing them down

ASME Section V Article 4 sets out the requirements of an ultrasonic examination procedure and splits them into essential and nonessential variables. On the essential side sit the things that change what the sound does: material and product form, thickness range, search unit type, frequency, angle and element dimensions, wedge or shoe, couplant, examination surface preparation and condition, scanning technique and overlap, the calibration block, the method of sizing, and — where scanning is mechanized or encoded — the scan increment, speed and software identification. Nonessential variables still belong in the document and still require a controlled revision when they change, but they do not put the demonstration in question.

The discipline this imposes is uncomfortable in a mill. Real inspection is adaptive: the technician arrives, the surface is worse than expected, the frequency comes down, a different wedge comes out of the case, and the job gets done. If the procedure's essential variables are written tightly enough to be meaningful, that adaptation is a deviation. If they are written loosely enough to permit anything, the procedure controls nothing. The resolution is a family of narrow procedures rather than one elastic one, with the technique choices pre-authorized for defined conditions.

This is also the point where thickness range is most often abused. A procedure claiming 0.200 in. to 6.000 in. of carbon steel, demonstrated on a single 1 in. block, has been demonstrated for a fraction of its claimed range. Auditors know it, and on plate and pipe work with genuinely wide thickness variation it is the first thing they check.

Damage mechanisms in steel and primary metals, and the technique each one forces

API RP 571 was written for refining, but its mechanism definitions travel. Oxidation and high-temperature sulfidation attack furnace and duct internals wherever fuel-fired heat is present. Carburization and metal dusting attack tube and reformer materials in carbon-rich, CO-bearing atmospheres, and metal dusting in particular is a real and expensive mechanism in direct-reduced iron plants. Creep and stress rupture govern hot blast mains, stove shells, hangers and any component living above roughly forty percent of its melting point. Thermal fatigue attacks expansion joints, duct transitions and anywhere thermal cycling meets restraint. Graphitization and spheroidization degrade carbon and C-½Mo steels held hot for years. Dew-point corrosion, sulfuric and hydrochloric, attacks gas cleaning trains downstream where the flue gas cools below its acid dew point.

Each of those produces a distinctive signature, and technique follows signature. General wall loss with an oxidized backwall wants low-frequency, dual-element thickness with an explicit rule for lost backwall echo. Localized erosion wants encoded mapping, because the mechanism produces a scallop narrower than the beam and a spot grid reports an average that is not the minimum. Creep and thermal fatigue want angle-beam coverage with defined skews, plus a near-surface strategy — a plain 45-degree shear beam has a genuine near-surface blind zone, and creeping-wave or tip-diffraction approaches exist for exactly that. Carburization changes the material itself, shifting velocity and making a fixed steel velocity assumption a systematic error rather than a random one.

A procedure that lists techniques without linking them to mechanisms is the recurring problem. The link belongs in the scope: this procedure is written to detect and size this mechanism, on this asset class, in this thickness range, and here is why the chosen frequency, angle and probe configuration are capable of it.

High temperature: probes, couplant and the errors that follow

Most conventional probes and wedges are rated for surface temperatures well under boiling. Above that, the procedure has to specify high-temperature probes with an appropriate delay line, a contact duty cycle — typically a few seconds on the surface, then off to cool — and a couplant rated for the temperature that will not carbonize into an acoustic barrier mid-scan. A procedure that says "high temperature couplant as required" has specified nothing, and the technician on the platform will make the decision instead.

Couplant carries a second constraint that steel plants encounter on austenitic and nickel components in gas cleaning and cryogenic service. ASME Section V limits residual sulfur in couplant used on nickel-base alloys and residual halides on austenitic stainless steels and titanium, on the order of a few hundred parts per million, because the residue promotes cracking after the inspector has gone home. That means a certificate of conformance per batch, retained with the examination record. Very few mills can produce one on request.

Then the arithmetic. Sound velocity in steel falls as temperature rises, so a gauge set with an ambient velocity over-reads on a hot surface by roughly one percent per hundred degrees Fahrenheit. On a 0.750 in. duct wall at 600°F, uncorrected, that is around 0.030 in. of phantom metal — enough to convert a real thinning trend into an apparent stable wall. Section V also caps the temperature differential between the calibration block and the examination surface, which a hot line defeats immediately. The procedure has to state the correction method, the block temperature control, and what happens when neither can be achieved.

Coarse grain, cladding and the material problems shear waves lose to

Centrifugally cast tube, heavy castings and austenitic weld metal all scatter and mode-convert ultrasound. At the sensitivity a procedure demands, the screen fills with grass, the signal-to-noise ratio collapses, and a technician either turns the gain down until the noise disappears — losing the flaws with it — or reports everything. A shear-wave weld procedure copied from wrought carbon steel practice is not merely suboptimal on these materials; it is incapable, and the failure is silent.

The available answers are known and belong in the document: lower frequency, larger element for penetration, transmit-receive longitudinal probes or dual-matrix arrangements that suppress the scatter common to both elements, and a realistic, written statement of the smallest flaw the technique can resolve in that material. That last item is what separates a professional procedure from a hopeful one. If the technique can resolve a 3 mm reflector in coarse cast material and no better, the procedure should say so, and the engineering that relies on it should know.

Clad and dissimilar metal welds carry an additional trap: distinguishing a genuine disbond from the interface geometry itself. A procedure that treats cladding as a footnote in a base-metal document will produce arguments during every outage. It needs its own calibration on a clad reference, its own criterion for what counts as disbond, and its own demonstration.

Calibration blocks, curvature and the constraints that get skipped

Section V constrains the calibration block on four axes simultaneously. Material: same product form, and acoustically similar in composition and heat treatment to the part. Thickness: the block thickness is tied to the nominal thickness of the part, with reflector depths at defined fractions of the wall. Curvature: for small diameters a curved block is required, and one curved block covers only a limited diameter band around its own size — roughly nine-tenths to one-and-a-half times the block diameter, which means a 12 in. block covers about 10.8 in. to 18 in. and nothing else. Temperature: the differential between block and examination surface is capped at a modest number of degrees.

In a mill, satisfying all four at once is the exception. Duct plate is a different heat treatment from the block on the shelf. Small-bore piping falls outside every curved block available. Hot work defeats the temperature constraint. The procedure's job is to say what happens then — which blocks are held, what their coverage bands are, what a technician does when the part falls outside them, and who authorizes the deviation. Silence produces improvisation, and improvisation is what an auditor discovers.

Calibration confirmation is the other half. Section V requires the system calibration to be confirmed at the beginning and end of each examination, on any change of examination personnel, and at intervals not exceeding four hours, with defined amplitude and sweep tolerances beyond which prior data must be re-examined or re-evaluated. Instrument linearity — screen height and amplitude control — carries its own periodic check regime under Article 5, with a screen-height check performed by setting reflectors to defined percentages of full screen height and verifying the second falls within tolerance. Both regimes belong in the procedure, and both are commonly missing.

What a qualification demonstration has to prove

Where the referencing code requires it, a procedure must be demonstrated to the satisfaction of the Level III, and the demonstration is a technical experiment, not a formality. It has to use a specimen containing flaws representative of the mechanism the procedure targets — not side-drilled holes standing in for creep cracking — placed at the depths, orientations and sizes that represent the least favorable condition the procedure claims to cover. It has to be run by a technician of the level the procedure names, following the written text with no verbal supplements.

What it must produce is evidence of four things: detection of the target flaws at the stated sensitivity; sizing within a stated tolerance, reported as an actual comparison against the true flaw dimensions rather than an assertion; coverage, meaning the scan plan and overlap demonstrably interrogated the required volume; and repeatability, ideally by a second operator on a separate occasion. Everything is retained — instrument settings, calibration records, scan plans, raw data, the flaw truth table.

Two failure modes recur. The first is a demonstration run by the procedure's author, who knows where the flaws are and unconsciously scans accordingly; blind or semi-blind demonstration is worth the extra effort. The second is a demonstration on ideal material that quietly establishes a capability the procedure then claims on coarse cast, clad or hot material where it was never shown. The demonstration's scope is the procedure's real scope, whatever the scope statement says.

Findings that recur when a mill UT procedure is audited

The list is remarkably stable across sites. The procedure claims a thickness or diameter range wider than the blocks on hand can cover. The couplant has no certificate of conformance for sulfur or halide content on austenitic and nickel components. Transfer correction is mentioned but never recorded on any report. Instrument linearity checks are absent or were last performed at purchase. Calibration confirmation intervals are not documented, so a shift's data cannot be bounded by a valid calibration on either side. The sizing method is unstated while heights are reported to three decimal places. Revision control shows the field carrying an obsolete copy.

Two more are specific to this industry. High-temperature velocity correction is either absent or applied inconsistently between technicians, which shows up as a trend line that moves with the season rather than with the corrosion. And structural weld examinations to AWS D1.1 are run with the dB rating arithmetic mishandled — the indication rating is built from the indication level, the reference level and an attenuation factor tied to sound path, and dropping or mis-signing the attenuation term changes accept and reject decisions on real welds. Anyone auditing an AWS report should recompute two ratings by hand; it takes five minutes and it is frequently productive.

None of these are exotic. They are the predictable consequence of procedures written once, filed, and never exposed to the assets they govern. Fixing them is a document exercise plus a small amount of hardware and demonstration effort, and it is far cheaper than defending a wall-loss trend that turns out to have been an artifact.

How Atlantis approaches the work

The engagement starts with a walkdown of the asset classes and the codes attached to them, then a read of whatever procedures exist. The output is a gap list separated into three tiers: findings that make existing data unreliable, findings that would fail a client or code audit, and housekeeping. In most mills the first tier is smaller than management fears and the third is larger than anyone wants to hear.

From there Atlantis writes the procedure family — typically thickness and corrosion mapping, weld examination, high-temperature scanning, and a dedicated document for cast or austenitic material — with essential variables written tightly, calibration hardware identified by what you actually hold, and each technique linked to the mechanism it exists to find. Where demonstration is required or advisable, the specimens, the flaw truth table and the acceptance of the demonstration are planned before the first probe goes on steel.

The procedures are approved by an ASNT Level III who remains available to defend them, revise them when the code edition or the plant changes, and answer technical questions during audits. Atlantis supplies NDT technical authority only — procedure development and qualification, written practice work, personnel certification within your practice, and independent review of inspection data. It is not the API inspector of record and not a PSM auditor. To scope a procedure package for your plant, contact info@atlantisndt.com; work is quoted on request.

Request a consultation