Writing and Qualifying UT Procedures for HRSGs, Headers and Outage Scope
An ultrasonic testing procedure for combined-cycle plant is qualified, not merely written: the essential variables in ASME Section V Article 4 fix what may change without re-demonstration, and the referencing code — Section I or B31.1 — sets acceptance. HRSG damage decides technique, and a procedure aimed at fusion-line flaws will miss Type IV creep cracking in Grade 91 girth welds.
Procedure development starts with the joint and the mechanism, not with a template. On a heat recovery steam generator the volume of interest is rarely a clean butt weld in flat plate: it is a stub-to-header connection with limited scan surface, an economizer elbow thinning by flow-accelerated corrosion on a smooth internal profile, or a Grade 91 girth weld where the damage sits in the fine-grained heat-affected zone a millimetre or two off the fusion line. Each of those drives a different search unit, a different scan plan and a different calibration reflector. The procedure then has to state the essential variables tightly enough that the qualification means something. A frequency range written as two to ten megahertz, or a wedge angle written as forty to seventy degrees, converts a demonstration on one configuration into a claim about dozens that were never tried — and that is the gap an auditor opens first.
Source: Written against ASME Section I and ASME B31.1 as referencing codes for HRSG pressure parts and power piping; ASME Section V Article 4 (including the Table T-421 procedure requirements and essential variable listing) and Article 23 for technique and thickness measurement; ASME Section IX where weld procedure and repair qualification interact; NBIC (NB-23) for in-service repair and alteration; API 579-1/ASME FFS-1 for fitness-for-service assessment of the results; API 571 for damage mechanism definition; ASNT SNT-TC-1A and ANSI/ASNT CP-189 for the personnel who execute the procedure.
| Component | Dominant mechanism | Technique the mechanism forces | Where a generic procedure fails |
|---|---|---|---|
| LP economizer and feedwater elbows, tees, downstream of orifices | Single-phase flow-accelerated corrosion | Close-grid thickness mapping indexed to a permanent datum so readings are trendable | A coarse spot grid steps over a smooth scalloped minimum and reports healthy wall |
| HP superheater and reheater header girth welds in Grade 91 | Type IV creep cracking in the fine-grained heat-affected zone | Angle beam or phased array aimed at the HAZ a short distance off the fusion line, proven on a Type IV mock-up | Sensitivity and scan angles tuned for fusion-line and volumetric flaws scan past the damage |
| Attemperator downstream piping and spray zones | Thermal fatigue from quenching during load changes and starts | ID-connected crack detection with a technique proven on thermal fatigue crack morphology | Amplitude criteria written for volumetric flaws under-call tight, branching cracks |
| Stub-to-header and tube-to-header connections | Thermal fatigue and creep at the weld toe, plus geometry stress | A limited-access scan plan with geometry-specific calibration and a defined coverage claim | Scan plan assumes two-sided access and full skip room that the tube bank does not allow |
| Dissimilar metal welds and buttered joints | Interface cracking with high attenuation and mode conversion | Low-frequency damped search units and a DMW-specific demonstration on representative material | Sensitivity established on a carbon steel block and applied across the interface |
| Steam drum and feedwater piping in older units | General and localised corrosion, plus original fabrication flaws | Thickness to Article 23 plus weld examination on a documented sample basis | Thickness survey substituted for weld examination because it is faster |
What the referencing code asks of a procedure, and what Section V does not decide
A common structural error in power-plant UT procedures is that they read as though ASME Section V were the whole requirement. Section V supplies method requirements — how the examination is performed, what the procedure must address, what calibration and verification are required — and Article 4 for welds and Article 23 for thickness measurement are where most combined-cycle scope lives. But Section V is explicit that acceptance criteria come from elsewhere. For HRSG pressure parts that is ASME Section I; for the balance of the steam and feedwater systems it is B31.1; where the work is in-service repair or alteration it is the NBIC, with the repair organisation's own accreditation and programme in play.
The consequence is that a procedure has to name its referencing code and edition and then live with that choice. Editions matter: acceptance criteria, personnel qualification references and permitted alternatives all move between editions, and a plant whose jurisdiction has adopted a particular edition cannot be examined against a different one because that is what the contractor's template said. When an authorised inspector or a jurisdictional boiler inspector reads a procedure, the edition line is usually the first thing checked and the easiest thing to get wrong.
There is a second layer specific to power. Results frequently feed a fitness-for-service evaluation under API 579-1/ASME FFS-1 rather than a straightforward accept-reject decision — a thinned economizer elbow assessed for continued operation to the next outage, a header ligament assessed for crack-like flaw acceptability. That downstream use changes what the procedure must produce. An assessment consumes dimensions, positions and a stated measurement uncertainty. A procedure that produces only accept-reject calls cannot feed it, and retrofitting the data afterwards means going back to the scaffold.
Essential variables, and the range trap
Section V Article 4 sets out, in tabular form, what an ultrasonic procedure must address and which of those items are essential — meaning a change beyond the range the procedure states invalidates the qualification and requires re-demonstration. The list covers the obvious mechanical items: search unit frequency, element size and nominal angle, wedge material and angle, couplant, surface preparation and condition, examination technique, the calibration block and reflectors, scanning technique and directions, and how sensitivity is established and verified. Nonessential variables can be changed under a procedure revision without a new demonstration.
The trap is not misunderstanding which variables are essential. It is writing the ranges so widely that the qualification loses meaning. A procedure that permits any frequency between 2 and 10 MHz, any angle between 40 and 70 degrees, and wall thickness from 6 mm to 100 mm has, on paper, been qualified for a configuration space of hundreds of combinations, of which one was ever demonstrated. Auditors know this. The question that opens the file is not whether the procedure lists essential variables but which specific combination the demonstration record covers, and whether the technique sheet in the field sits inside it.
There is a discipline that avoids the argument entirely: write the procedure to the scope you actually run, then add technique sheets that pin down the exact combination for each application, each traceable to a demonstration. A narrow procedure with a family of qualified technique sheets is stronger, easier to audit and easier for a technician to execute correctly than a wide procedure that promises coverage nobody proved. It also makes the eventual scope expansion a defined piece of work rather than an unnoticed drift.
Damage mechanisms choose the technique, not the other way round
Procedure development for gas and combined-cycle plant should begin with API 571-style mechanism identification for the specific circuit, because the mechanism dictates the flaw morphology, the flaw's location and orientation, and therefore the technique. Combined-cycle units concentrate a distinctive set. Flow-accelerated corrosion thins the low-pressure economizer and evaporator circuits and feedwater piping, favouring elbows, tees and the region downstream of flow disturbances, and it does so with a smooth, scalloped internal profile rather than discrete pitting. Thermal fatigue attacks attemperator downstream piping and stub-to-header connections, driven by the temperature swings that two-shifting a unit imposes several hundred times a year.
Creep damage concentrates in the high-temperature headers and steam leads, and in Grade 91 material it takes a specific and awkward form discussed in the next section. Dissimilar metal welds, where they exist, crack at the interface. Older units carry general corrosion and original fabrication flaws in drums and large-bore piping. Each of those has a different answer to the questions that build a technique: where is the flaw, which way is it oriented, is it surface-connected or embedded, and what does the surrounding material do to sound.
Getting this wrong produces a procedure that is technically correct and operationally useless. Flow-accelerated corrosion is the clearest illustration. It is a thickness problem, but a spot-reading grid sized for general corrosion — readings every 50 or 100 mm — will straddle a scalloped minimum and report comfortable wall on a component approaching its limit. The mechanism forces close-grid or encoded mapping, and it forces the grid to be indexed to a permanent physical datum so that the next outage measures the same points and the trend means something. Untrended thickness data on an FAC circuit is a cost with no information content.
Grade 91 and Type IV cracking: why a competent generic procedure still misses it
Creep strength enhanced ferritic steels, Grade 91 above all, deliver their properties through a specific tempered martensitic microstructure, and that structure is disturbed by the weld thermal cycle. In the outer heat-affected zone the material is heated just above transformation and forms a fine-grained region that is measurably weaker in creep than either the base metal or the weld metal. Under sustained high-temperature service, damage accumulates there — Type IV cracking — typically initiating subsurface, running roughly parallel to the fusion line, a short distance outside it.
That geometry defeats a scan plan built for conventional weld flaws. A procedure that sets sensitivity and angles for lack of fusion, incomplete penetration and volumetric flaws is aiming at the fusion line and the weld volume. Type IV damage sits outside that target, in coarse material that attenuates and scatters, and it presents at low amplitude before it is large. Crews with excellent technique report clean welds because the technique was aimed elsewhere. This is not a competence failure; it is a procedure failure, and it is why Grade 91 header scope needs its own procedure rather than a note appended to the general weld procedure.
Doing it properly means three things. The scan plan has to interrogate the HAZ specifically, with angles and beam paths chosen to intersect flaws parallel to the fusion line rather than normal to the scan surface. The demonstration has to be performed on a specimen containing Type IV damage or a credible representation of it — flat-bottom holes and side-drilled holes will not stand in for a diffuse creep-damaged zone. And the procedure should state its detection threshold honestly, because early creep damage is a cavitation field before it is a crack, and metallographic replication remains a complementary tool rather than something ultrasonics replaces. A procedure that quietly implies UT alone will catch damage at any stage is overselling the method.
The qualification demonstration and what it has to prove
A qualification demonstration exists to convert a written claim into evidence, and its design is where most of the technical thought belongs. Representativeness is the whole test: material grade and product form, wall thickness, weld configuration and access, surface condition, and where relevant the presence of cladding, buttering or a dissimilar interface. A demonstration on machined plate proves that the equipment functions and the operator can drive it. It does not support a claim about a stub weld surrounded by tubes on a header at height.
What the demonstration must show depends on what the procedure claims. If the claim is detection, it must show detection of flaws of the type and size that matter, at the sensitivity the procedure establishes, with the scan coverage the technique sheet specifies — including a coverage statement where geometry genuinely prevents full-volume examination. If the claim extends to sizing, it must show sizing within a stated tolerance, because a fitness-for-service assessment will consume those dimensions as though they were measurements. And it must be executed by personnel of the level who will perform the work, not by a specialist who will not be on site.
In non-nuclear power, the qualification burden is set by the referencing code plus whatever the owner imposes, and many utilities and independent power producers impose mock-up demonstrations of their own on high-consequence scope. Treat that as an asset rather than an obstacle. An owner-witnessed demonstration on a representative mock-up settles the coverage argument permanently, gives the contractor a defensible record across the fleet, and converts what would otherwise be an outage-floor negotiation into a document that was signed months earlier.
Temperature, and the arithmetic that turns non-conservative
Outage schedules push examinations onto components that have not fully cooled, and temperature attacks ultrasonic measurement from three directions at once. Sound velocity in steel falls as temperature rises. Wedge material properties change, which shifts the refracted angle and the incident point of an angle beam probe. And standard couplants break down, changing coupling efficiency mid-scan without any obvious indication on the screen.
The velocity effect produces the arithmetic trap that matters most, because it runs the wrong way. Thickness is computed from transit time and an assumed velocity. On a hot component the true velocity is lower than the room-temperature value the instrument holds, so the true transit time is longer, and the instrument — dividing by a velocity that is too high — returns a thickness greater than the actual wall. The error is roughly one percent per 100 °F above ambient as a working rule of thumb, so a component at 300 °F can read around two percent thick. On a 12 mm economizer elbow near its minimum, that is a quarter of a millimetre of wall that does not exist, reported in the conservative direction's opposite.
A procedure has to close this explicitly: state the temperature range within which the technique is qualified, require high-temperature couplant and wedges rated for the range, and either mandate a velocity check against a heated reference of the same material or specify the correction and require it to be recorded on the report. It should also require the component temperature to be measured and written down, not estimated. Reports that carry no temperature field cannot be corrected retrospectively, which means a whole outage's readings become uncertain the moment somebody asks how hot the line was.
Findings that recur when the procedure is audited
Audit findings on power-plant UT procedures cluster tightly. The procedure references a superseded code edition, or cites Section V without ever naming the referencing code that supplies acceptance criteria. The essential variable ranges are wider than the demonstration record. The calibration block is not of representative material, product form or heat treatment, and no transfer correction is recorded on reports even though the procedure requires one. Instrument linearity verifications are performed at the required intervals but retained without identifying the instrument or the technician.
On the execution side: scan plans that do not account for the actual joint geometry, with the coverage limitation neither calculated nor stated; technique sheets in the field at revisions the controlled procedure never issued; recording thresholds applied inconsistently between crews; encoded data files retained without the setup parameters needed to interpret them later. Data retention is a quiet one — an encoded scan whose configuration file has been lost is a picture nobody can re-evaluate, which matters enormously when the same weld is examined again in three years and somebody wants to know whether an indication grew.
And on personnel: technicians certified in UT generically and deployed on encoded corrosion mapping or phased array without a limited certification or documented technique qualification; contract crews working under certificates the plant filed without any written acceptance route; practical examination records that show a score and no specimen. These findings do not make the examination wrong. They make it unsupportable, which in a dispute over a repair decision amounts to the same thing.
Writing to the outage calendar rather than during it
Outage economics dominate everything about power-plant NDT. A combined-cycle unit out of service is a revenue loss per day that dwarfs the inspection budget, so the schedule compresses and anything that lands on the critical path gets negotiated rather than solved. Procedure qualification is the classic critical-path item: a technique that has not been demonstrated when the scaffold goes up will either be demonstrated under pressure or waived under pressure, and both outcomes are visible in the record afterwards.
The alternative is to treat procedure development as running-plant work. Scope forecasting from the previous outage's findings and the plant's own mechanism review identifies which examinations will be required. Anything needing qualification is identified early, specimens are procured or fabricated, demonstrations are run and witnessed while the unit is generating, and technique sheets are issued and revision-controlled before mobilisation. The outage then executes qualified techniques rather than inventing them.
This also fixes the personnel problem in advance. Surge crews arriving for a three-week outage need certifications your written practice actually accepts and technique-specific qualifications for anything beyond manual UT. Both take time to arrange and neither can be arranged on day one of an outage. Plants that plan the paperwork on the same calendar as the scaffolding lose noticeably fewer hours to it. To discuss procedure development, qualification design or an outage readiness review, request a consultation at info@atlantisndt.com.
Which code does a power-plant UT procedure actually have to satisfy?
Two layers. Section V provides the method requirements — Article 4 for welds, Article 23 for thickness measurement — but Section V does not decide acceptance. The referencing code does: ASME Section I for boiler pressure parts, B31.1 for power piping, and NBIC where the work is a repair or alteration in service. A procedure that cites Section V alone and never names its referencing code and edition has left its acceptance criteria undefined.
What counts as an essential variable, and why does the distinction matter?
An essential variable is one whose change beyond the range stated in the procedure invalidates the qualification and forces re-demonstration — search unit frequency, size and angle, wedge and couplant, surface condition, calibration block, scanning technique and directions, and the examination sensitivity basis. Nonessential variables may be changed with a procedure revision but no new demonstration. The practical significance is that essential variables define exactly how much of your real scope the demonstration you performed actually covers.
Why does Grade 91 change the whole technique discussion?
Because the damage does not sit where a conventional weld scan looks. Type IV creep cracking initiates in the fine-grained heat-affected zone just outside the fusion line, often subsurface, in a material whose coarse structure attenuates and scatters. A scan plan optimised for fusion-line lack of fusion and for volumetric weld flaws can pass over it repeatedly. The technique has to be aimed at the HAZ specifically and proven on a specimen containing that morphology.
What must a qualification demonstration actually show?
That the technique, executed by the people who will execute it, finds the flaws of interest in material, thickness, geometry and surface condition representative of the component — and, where sizing is claimed, that it sizes them within a stated tolerance. A demonstration on a flat plate calibration block with side-drilled holes proves the instrument works. It proves nothing about a stub weld in a congested tube bank, which is where the examination will happen.
Is API 510, 570 or 653 inspector training part of this offer?
No. Those are API inspector certifications, and they authorise a named person to sign in-service inspection of vessels, piping and tanks. What is supplied here is procedure development and qualification, the technical authority behind it, personnel certification within your written practice, and independent review of the resulting data. The inspector of record and the engineering acceptance decision remain with your organisation.
How far ahead of an outage should procedure work start?
Far enough that no demonstration lands on the critical path. Drafting is fast; qualification is not, because it needs representative specimens, instrument time and the same personnel who will work the outage. Plants that start a season ahead qualify calmly and mobilise with technique sheets ready. Plants that start at mobilisation discover the coverage gap on the scaffold, negotiate a deviation under schedule pressure, and inherit a finding at the next audit.