Steel Structure Inspection and NDT: A Practical Field Guide

Non-destructive testing (NDT) is where the quality promise of a steel structure contract is actually kept. A weld can look perfect and contain a crack that will grow under fatigue; a plate can pass the mill certificate and hide a lamination that turns into lamellar tearing at a T-joint. The purpose of inspection is not to be thorough for its own sake — it is to catch, quantify and allow repair of discontinuities that the design load assumptions cannot tolerate. On a long-span grid or truss project, the inspection and test plan (ITP) decides, weld by weld, how much evidence the final file must contain.

This guide covers the four techniques that dominate structural steel inspection — visual (VT), ultrasonic (UT), magnetic particle (MT) and radiographic (RT) testing — their detection limits and blind spots, the acceptance levels that govern them, the typical testing ratios used on steel structures, and how a third-party inspection workflow (SGS, TÜV, Intertek or client-appointed bodies) actually operates at a fabrication plant. It is written from the fabrication shop floor, where inspection is a production discipline rather than an administrative ritual.

1. Visual Testing (VT): The 100% Gate

Visual testing per ISO 17637 is the first and cheapest line of defence, and it is applied to 100% of welds on every serious project. A certified welding inspector (to EN ISO 9712 or AWS CWI) checks weld geometry against ISO 5817 classes — excess weld metal, undercut depth, weld profile, root concavity, surface porosity and cracks visible to the eye or under 10× magnification. Weld gauges and fillet weld gauges quantify profile; crack-detection dye penetrant (PT) supplements visual for tight surface openings.

VT has hard limits: it finds surface-breaking and near-surface defects only, and its probability of detection drops sharply below about 0.5 mm crack width. It is the gate because it is fast, immediate and zero-cost per joint — the shop knows immediately whether the weld topping is out of tolerance, long before any volumetric testing. In our workflow, every weld receives VT before it is eligible for NDT sampling; a joint that fails VT is repaired and re-inspected before it enters the volumetric queue.

2. Ultrasonic Testing (UT): The Workhorse for Butt Welds

Ultrasonic testing per EN ISO 17640 (the GB/T 11345 family is the Chinese counterpart, technically equivalent in principle) is the standard method for full-penetration butt welds in structural steel. A probe sends a 2-5 MHz pulse into the plate; reflections from internal discontinuities are displayed on an A-scan with a distance-amplitude-correction (DAC) curve that normalises signal amplitude against depth. Angled probes (45°, 60°, 70°) are swept in defined patterns on both sides of the weld, and the technician records indication position, length and amplitude.

Acceptance follows ISO 11666 for the amplitude/length evaluation, or ISO 23279 for characterisation when needed; on our projects the common acceptance level for primary full-penetration welds is ISO 11666 level 2 (the companion to ISO 5817 level B/C welding classes), which rejects individual indications that exceed the length and amplitude limits for the thickness band — in the 10-40 mm thickness range, for example, single indications longer than roughly 10 mm at recording amplitude are cause for evaluation or repair. Three practical points distinguish a good UT programme:

• Coverage is directional: each probe angle covers a particular flaw orientation; planar defects (lack of fusion, cracks) are missed if the probe angle is unfavourable, which is why multiple angles and two-sided scanning are specified for critical joints.
• Geometry limits UT: coarse-grained austenitic welds scatter the beam, and weld caps that have not been ground interfere with probe coupling — the ITP must define whether the cap is ground and which joints are exempt.
• Personnel qualification decides detection probability: UT is operator-dependent; certificated level 2 technicians (EN ISO 9712 or equivalent) are a contractual requirement on all EXC3 and most EXC2 programmes.

3. Magnetic Particle Testing (MT): Surface and Near-Surface Cracks

Magnetic particle testing (ISO 17638 / GB/T 15822) detects surface and shallow subsurface discontinuities in ferromagnetic steel. The area is magnetised with a yoke (lifting force checked at 2-4 kN per the standard), sprayed with black or fluorescent particles, and indications — the particle build-up at crack openings — are evaluated under white light or UV-A (fluorescent). Acceptance is per ISO 23278, commonly level 2X: no linear indications greater than a defined length (typically 2-3 mm for cracks at weld toes), no > 3 mm rounded indications in a cluster, and no cracks whatsoever.

MT is the method of choice for fillet weld toes, weld roots after back-gouging, flame-cut edges and repaired areas — exactly the places where fatigue cracks initiate. It is fast (minutes per joint), sensitive to tight surface cracks that VT cannot see, and cheap, which is why it is the default sampling method for fillet welds on critical connections. Its limitation is depth: anything more than 1-2 mm below the surface is out of reach, and demagnetisation is required on members near magnetically sensitive equipment.

4. Radiographic Testing (RT): Film and Digital

Radiographic testing per ISO 17636-1 (film) or ISO 17636-2 (digital CR/DR) projects an X-ray or gamma image of the weld onto a detector. It gives a permanent, reviewable record of volumetric condition — porosity, slag, tungsten inclusions and gross lack of fusion are clearly readable — and its detection of volumetric flaws is superior to UT. Source selection follows thickness: selenium-75 up to roughly 40 mm, iridium-192 up to 60 mm, and cobalt-60 for 60-200 mm; X-ray equipment is preferred in plants where safety permits. Image quality is verified with image quality indicators (IQI) per ISO 19232.

RT’s trade-offs are cost, safety and the directional blind spot: planar defects oriented perpendicular to the beam (transverse cracks) can be missed unless supplemental shots are planned, and radiation protection zones slow production in busy shops. In practice, our plants use RT where the project or class demands it — thick full-penetration joints where surface geometry suits film, T/Y-joints in heavy trusses, and as the specified verification method on some EXC3/EXC4 members — while UT remains the daily volumetric workhorse. The ITP, not habit, decides which.

5. Weld Classes and Typical Testing Ratios

Testing ratios are defined by the project specification, the design standard and the execution class. On Chinese-executed steel structures the GB 50205 practice is a sensible reference; on EN 1090-2 projects the ratios follow the execution class and the ITP. A typical mapping used on our shop floor:

Weld category	Typical NDT	Typical proportion	Acceptance standard
All welds (every programme)	VT (ISO 17637)	100%	ISO 5817 level B/C/D per execution class
Full-penetration butt welds in primary members	UT (ISO 17640 / GB/T 11345) or RT	100% for first-grade structures; not less than 20% where code permits sampling	ISO 11666 level 2 / ISO 10675-1
Fillet welds and partial penetration welds at critical connections	MT (ISO 17638), PT on non-ferrous or inaccessible	10-20% sampling typical; 100% on tension-critical weld toes where specified	ISO 23278 level 2X
Weld roots after back-gouging (full-penetration)	MT or PT on gouged surface	100% before sealing pass	No cracks, no laminations at root surface

When sampling is allowed, the sample is taken at the joints the fitter regards as highest risk — first piece of a shift, changes in plate thickness, overhead positions — and any failure expands testing to 200% of the original sample, including all joints made by the same welder in the same period. This escalation rule is written into our ITP and is more valuable than any fixed ratio.

6. Third-Party Inspection: How SGS, TÜV and Intertek Work

Most international contracts appoint a third-party inspection body to witness and verify the fabricator’s quality system. The workflow is anchored on the ITP, which lists every inspection activity, its acceptance standard and its witness type:

• Witness point (W): the inspector is invited and the activity proceeds on schedule; the inspector signs off retrospectively or attends when available.
• Hold point (H): the activity stops until the inspector attends, verifies and releases it. Typical hold points: material certificate review, WPS/WPQR and welder qualification audit, weld joint set-up before welding, NDT witness, coating DFT, final dimensional release.
• Review point (R): documents are submitted for review without blocking production — NDT reports, coating records, dimensional reports, packing lists.

A realistic third-party cycle for a large structure spans: pre-production meeting (ITP confirmation, hold-point calendar), material stage inspection (mill certificates, UT on thick plates), fabrication surveillance (welder audits, joint inspection, NDT witnessing), coating stage (surface preparation, DFT measurement per ISO 19840), final inspection (dimensions, marking, packing list against drawings), and release — at which point the body issues either a release note or an EN 10204 3.2-style countersigned certificate for the items stipulated in the contract. Scheduling these visits against the fabrication programme is a production-planning task in its own right; a shop that treats third-party visits as disruptions, rather than milestones, is telling you how it will behave on site.

7. Common Defects and Their Prevention

Defect	Typical cause	Detected by	Prevention
Gas porosity (pores, clusters)	Damp flux or wire, contaminated surface, shielding gas disruption, excessive travel speed trapping gas	VT (surface), RT/UT (internal)	Dry and bake consumables per data sheet; clean joint; check gas flow and wind shielding
Slag inclusions	Incomplete interpass cleaning, wrong travel angle, low current leaving slag trapped at toes	RT/UT (internal), MT rarely	Full grinding between passes; correct current and angle per WPS
Lack of fusion / lack of penetration	Low heat input, poor joint fit-up, wrong root gap, off-angle arc	UT/RT (linear, planar indications)	Fit-up verification before welding; WPS-compliant parameters; root pass inspection
Cracks (hot, cold, hydrogen)	High restraint + hydrogen, insufficient preheat, fast cooling (hardened HAZ), sulphur-rich parent metal	MT/PT (surface), UT (sub-surface); often delayed 24-72 h	Preheat control, low-hydrogen consumables (H4-H5), Ceq limits, post-weld hydrogen diffusion for constrained joints

Two rules govern defect management on our projects. First, timing: hydrogen-induced cold cracks can appear 24-72 hours after welding, which is why critical welded connections are held for a minimum delay before NDT — the ITP schedules UT/MT no earlier than 24 h after weld completion. Second, repair discipline: a crack is never simply re-welded over; the defect is removed to sound metal, verified by MT, welded with a qualified repair WPS, and re-tested 100%. Repair history is documented per member and reported to the third party, because the entire point of the inspection file is that the next engineer can trust it.

Key Takeaways

• VT is the 100% gate, never the last word. Visual testing catches profile and surface defects on every weld; volumetric and surface methods catch what VT cannot see.
• UT is the volumetric workhorse for butt welds (ISO 17640/GB/T 11345, acceptance ISO 11666 level 2), with 100% coverage typical on primary full-penetration welds and 10-20% MT sampling on fillet welds.
• Every technique has a blind spot. UT struggles with coarse grain and unground caps; RT misses planarly oriented defects and costs more; MT sees only the near-surface. The ITP, not the cheapest option, decides the method.
• Third-party inspection runs on hold points. SGS, TÜV and Intertek operate through an ITP with witness, hold and review points — factor their visits into the production schedule, not around it.
• Timing and repair discipline separate good programmes from paperwork. Critical welds are NDT-ed after a 24-72 h delay for hydrogen cracks, and repairs are documented, re-qualified and re-tested 100%.

Our plants run full in-house NDT (VT/UT/MT/RT) with EN ISO 9712-certified personnel and accommodate third-party inspection at every hold point; our manufacturing qualifications and inspection practice are documented separately. If you are setting up an inspection programme for a steel project — or auditing a supplier’s NDT file — contact our engineering team for a review of the ITP, acceptance levels and hold-point schedule before fabrication starts.