NDT Test for Welding: Methods, Codes, and How to Choose

The weld looked perfect. The beads were uniform, the toes were clean, and the fabricator had already moved on to the next joint. Eighteen months later, that same joint was being cut out with a carbon arc gouge after a routine inspection found lack-of-fusion running along a third of its length. Someone had looked at that weld. Nobody had actually tested it.

Why an NDT Test for Welding Matters More Than the Weld Itself

An NDT test for welding is a formal examination of a completed or in-progress weld that reveals discontinuities without cutting, sawing, or destroying the part. NDT stands for nondestructive testing, and you will also see it written as NDE, nondestructive examination, or NDI, nondestructive inspection. All three mean the same thing in practice: a trained person applies a proven physical method, judges the result against written acceptance criteria, and records what they found. The joint stays whole and, in most cases, stays in service.

That is the clean definition. The reason it matters is less clean. A weld can be geometrically perfect on the outside and still be unsound on the inside. Lack of fusion can track along a sidewall. A crack can sit a millimeter below the cap. Porosity can cluster at the root where nobody will ever see it. None of these announce themselves. They wait for load cycles, thermal expansion, pressure, vibration, and time.

This is why inspection and testing are two different activities. Inspection is somebody looking. Testing is a method with a written procedure, calibrated equipment, a qualified operator, and a defined accept/reject line. A fabricator who “looked at it” has done neither.

What NDT Actually Sets Out to Find

The target of every NDT method is a discontinuity — an interruption in the expected structure of the material. That word choice is deliberate. AWS B1.10M/B1.10:2016, the Guide for the Nondestructive Examination of Welds, is careful to separate a discontinuity from a defect. A discontinuity simply exists. A defect is a discontinuity that exceeds the acceptance criteria for that specific weld. The same porosity cluster that is a rejectable defect in a cyclically loaded bridge detail may be perfectly acceptable in a lightly loaded bracket.

That distinction is the hinge the whole discipline swings on, and AWS B1.10 makes the point bluntly: a sentence like “the weld shall be radiographically examined” is incomplete. It tells you nothing until it also names the acceptance standard. Examined to what limit? Judged by whose criteria? Without that answer the test is theater.

So NDT is not a search for perfection. It is a comparison between what is actually in the weld and what the governing document says is tolerable. Anyone who tells you the goal is “zero defects” has not read the code they are quoting.

The Seven Methods in AWS B1.10

The AWS guide catalogues seven examination methods for welds. You do not need to master all seven to make good decisions, but you do need to know what each one can and cannot see, because the most expensive NDT mistake in the industry is choosing the wrong method for the defect you are worried about.

• Visual testing (VT) — direct or remote observation of surface condition, geometry, and workmanship. The most common method in structural steel fabrication, according to the American Institute of Steel Construction, and per AWS B1.10 typically the least expensive. Its limit is absolute: surface only.
• Liquid penetrant testing (PT) — a liquid drawn by capillary action into surface-breaking openings, then developed out as a visible indication. Works on almost any clean, nonporous material.
• Magnetic particle testing (MT) — magnetization of a ferromagnetic part, with iron particles pulled to leakage fields at surface and near-surface discontinuities. Fast and sensitive, but blind to austenitic stainless and aluminum.
• Radiographic testing (RT) — X-rays or gamma rays passed through the weld, with differential absorption recorded on film or a digital detector. Excellent for volumetric flaws, weak on tight planar cracks oriented the wrong way.
• Ultrasonic testing (UT) — high-frequency sound introduced at an angle, with echoes interpreted to locate and size reflectors. The strong choice for planar defects in thicker sections.
• Electromagnetic or eddy current testing (ET) — induced currents in conductive material, disturbed by surface and near-surface discontinuities. Common on tube and overlay work.
• Leak testing (LT) — bubble immersion, pressure decay, helium mass spectrometry, and similar techniques that prove, or disprove, containment.

You can read the source document directly through the AWS B1.10 guide preview if you want the full treatment; it is short, plain-spoken, and far more readable than most standards. The broader catalogue of test methods lives in the ASTM nondestructive testing standards collection, which covers the individual method practices in detail.

The Economic Logic: Failures Don’t Happen at the Shop

Hidden flaws are cheap to find and expensive to ignore. The cost curve is brutal and predictable. A discontinuity caught at the bench costs a grind and a re-weld. Caught after final inspection, it costs a re-test. Caught after paint, it costs blasting, re-testing, repainting, and a schedule hit. Caught in service, it can cost a shutdown, a recall, an injury, or a fatality — plus the investigation, the legal exposure, and the reputation.

That is the whole business case, and it is why NDT exists as a separate trade rather than a line item on a welder’s checklist. The test is not there to grade the welder. It is there to move the discovery of a problem to the cheapest possible moment.

There is a second logic, and it is arguably more important: safety. Pressure vessels, lifting gear, structural moment frames, marine hulls, and process piping do not fail politely. They fail when people are standing near them. NDT is one of the few engineering controls that catches a flaw while it is still a flaw instead of an incident.

What This Guide Covers, and Who It’s For

This is not a textbook recital of method theory. What follows is a working decision framework: what each method physically does, what it misses, what the codes require, what it costs, and how to avoid the mistakes that turn a real inspection into a rubber stamp.

It is written for the people who actually have to make these calls:

• Welders who want to know what the inspector is doing and why, so the next joint passes the first time.
• Fabricators and shop supervisors balancing throughput against a testing clause nobody read closely.
• Junior QC staff handed a report template and a deadline.
• Engineers who inherited an inspection specification from a previous project and are not certain it still applies.
• Hobbyists and small shops shipping a pressure part, a boat hull, or a structural repair who need to know when to call a professional instead of guessing.

Because material choice drives method choice so heavily, a few of the practical examples later in this guide touch on aluminum work — the kind of thing covered in our guide to aluminum MIG welding, where surface preparation and oxide removal matter as much to the test result as to the weld itself. The same principle applies across processes: what you can test, and how reliably, depends on what you built it from.

One more thing before the methods. NDT cannot rescue a badly designed joint, and it cannot rescue a process that was never qualified. Testing sits at the end of a chain that starts with a welding procedure specification and a qualified welder. The stronger that chain, the less work the test has to do — and the more likely the test result is meaningful rather than merely survived.

NDT vs. Destructive Testing: The Vocabulary That Decides Whether a Weld Passes

Nondestructive testing doesn’t exist in a vacuum. Every NDT method used on a production weld is backed by destructive testing performed upstream, usually months earlier, on a coupon that no longer exists. Understanding that relationship separates a fabricator who argues about inspection results from one who can resolve them.

Destructive Testing: The Half You Never See on the Job

Destructive testing cuts, pulls, bends, or fractures the specimen. It destroys the thing it measures, which is exactly why it can measure properties NDT can only infer:

• Tensile test — pulls a coupon until it breaks, reporting yield strength, ultimate strength, and elongation, proving the filler metal and procedure deliver the strength the design assumed.
• Guided bend test — bends a root, face, or side coupon through a specified radius. A ductility and fusion check that is brutal on root lack of fusion.
• Macro etch — sectioned, polished, and acid-etched to reveal penetration, fusion line, pass sequence, and heat-affected zone profile.
• Charpy V-notch — measures absorbed impact energy at a specified temperature, which matters enormously for low-temperature service and toughness-critical steels.
• Hardness survey — maps hardness across the weld and HAZ, flagging regions likely to crack in service or exceed a code ceiling.

The division of labor is simple. Destructive testing proves the recipe; nondestructive testing proves the meal. A procedure qualification tells you the combination of process, filler, preheat, and parameters can produce a sound weld. It says nothing about whether the joint in front of you actually got that treatment.

The Vocabulary That Decides Whether a Weld Passes

Most inspection disputes are not technical. They’re linguistic. These eight terms carry the weight:

• Discontinuity vs. defect. A discontinuity is any interruption in the weld’s intended structure — porosity, slag, undercut, a stray arc strike. A defect is a discontinuity that exceeds the acceptance criteria. Welders who use the words interchangeably end up grinding out indications that were never rejectable.
• Indication vs. flaw. An indication is the signal the method produced — a dark spot on a radiograph, a paint bleed-out, an echo on a screen. A flaw is a real, confirmed imperfection. Indications require interpretation; some are geometry, some are film artifacts. ASTM E1316 is the terminology standard inspection guidance points to when these words need to mean something precise.
• Acceptance criteria. The written limits that define pass and fail: maximum porosity size, allowable undercut depth, crack prohibition. Without them, an inspector has an opinion, not a verdict.
• Code vs. specification vs. standard. A code is a body of mandatory rules adopted into law or contract (ASME BPVC, AWS D1.1). A specification states project-specific requirements. A standard is a published consensus document that codes and specifications reference.
• Written procedure. The document describing exactly how the test is performed. AWS B1.10 is explicit that all NDE methods must be performed to an approved written procedure, that this is almost always a code requirement, and that documented procedures are what make results reliable and reproducible.
• Technique sheet. The job-specific version of the procedure — this joint, this thickness, this angle, this scan pattern, this technician.
• Calibration. Verified equipment response against a known reference, with records. Expired calibration is the fastest way to invalidate an otherwise perfect inspection.
• Probability of detection. No method finds every flaw. Each has a detection likelihood that varies with flaw type, orientation, size, and access. “100% inspection” means 100% coverage, not 100% certainty.

Why “Zero Defects” Is Not an Acceptance Standard

Every fusion weld contains discontinuities, and some are unavoidable at a given thickness, process, and position. That’s why codes set size limits rather than demanding perfection. An inspector cannot accept or reject a weld without a written criteria document in hand, so ask for it before the first arc is struck. If a customer says “just make it perfect,” get that translated into numbers on the drawing or in the specification. AWS A2.4 exists precisely so method, extent, and frequency can be shown unambiguously instead of negotiated later.

Where Procedure Qualification Ends and Production NDT Begins

This is the confusion that costs the most money. In the ASME world, Section IX covers qualification of welding procedures and personnel — the PQR, the WPS, the welder performance test. Section V covers how NDT is actually performed. Different documents, different questions.

AWS D1.1 follows the same logic: procedure qualification demonstrates a joint detail and process combination can meet requirements, while the inspection clauses govern production welds. Passing a procedure qualification never certifies a production weld, and NDT of a production weld never qualifies a procedure. The principle holds whether you’re qualifying a structural groove weld or dialing in a process like aluminum MIG welding, where procedure variables shift faster and matter more.

Get the sequence right: qualify the procedure, qualify the welder, produce the weld, then test it nondestructively against written criteria. Anything out of order is guesswork dressed as quality control.

Visual Testing (VT): The Cheapest Test You Are Probably Doing Badly

Ask a room of welders what NDT means and most will say radiography or ultrasonic. Visual testing rarely makes the list, which is odd, because VT decides the fate of more welds than every other technique combined. It is also the one most often performed by whoever walks past the bench with a flashlight and an opinion.

That gap between VT as a formality and VT as an engineered inspection is where money gets lost. AWS B1.10, the guide to nondestructive examination of welds, lists visual examination as usually the least expensive NDE method available, while noting it is limited to surface discontinuities. Cheap and limited is not the same as unimportant. Every volumetric method in this article inspects a weld that VT already had a chance to reject.

VT Is a Code Requirement, Not a Courtesy

The AISC solutions center on welding inspection and NDE makes a point that surprises people who assume NDT means UT or RT: AWS D1.1/D1.1M requires all welds to be visually inspected. Not a sample. Not the ones that look suspicious. All of them. Radiography and ultrasonic testing cover a subset of welds under specific clauses; VT is the floor beneath everything else.

AISC Specification Chapter N splits visual inspection into tasks before, during, and after welding. That structure is the whole idea in miniature. Treat VT as a single post-weld glance and you have done roughly a third of the job.

The Three Phases and Why the First Two Matter Most

Before welding is where VT has no competition. It is the only method that can verify joint fit-up, root gap, and interpass condition before a flaw is buried under fifteen passes of filler. Nobody can radiograph a root gap that closed during tacking, and no ultrasonic beam will report that mill scale was never removed. Pre-weld VT catches wrong consumables, wrong bevel angles, contaminated faying surfaces, and misaligned members while they are still cheap to fix. Cleanliness belongs here too; on aluminium work, surface prep is a genuine inspection item, and the guidance in aluminum cleaner for welding matches what an inspector looks for at the joint.

During welding covers interpass cleaning, pass sequence, travel-speed red flags, and root condition before the next layer hides it. On multi-pass work, a slag line left between passes becomes an internal planar flaw that only UT will find, at ten times the cost of catching it with a chipping hammer and a light.

After welding is gauge work: weld size, throat, length, undercut depth, convexity, overlap, spatter, arc strikes, and crater condition.

What VT Can and Cannot See

VT reads surface geometry and surface-breaking discontinuities. It reliably finds undercut, excessive convexity or concavity, insufficient leg, overlap, crater cracks, visible porosity, misalignment, and spatter. It cannot find lack of fusion, internal porosity, slag inclusions, or a crack that never reaches the surface. A tight toe crack under paint stays invisible until the coating is removed or a surface method is applied. That limitation is exactly why codes pair VT with PT, MT, RT, or UT instead of replacing them.

Tools, Lighting, and Access

• Weld gauges — fillet leg and throat, undercut depth, reinforcement height, misalignment.
• Fillet and undercut gauges — the two measurements most often eyeballed and most often wrong.
• High-intensity lighting — codes specify minimum illumination and it is measured, not guessed. A phone torch fails the requirement and the intent.
• Borescopes and remote visual inspection — for pipe internals, box sections, and tank interiors; the recordings become part of the record.
• Magnification and mirrors — useful at toes and in tight corners where the viewing angle is wrong.

Access matters as much as equipment. If the inspector cannot reach the code-specified viewing distance and angle, the inspection did not happen, whatever the report says. Scaffolding and lighting are inspection costs, not site overhead.

Timing: Cool First, Then Wait

Post-weld VT must wait until the weld has cooled enough to be examined safely and accurately. For crack-susceptible steels — thick sections, high-restraint joints, high-carbon equivalents, some Q&T grades — a deliberate delay before final visual examination is standard practice, commonly 24 to 48 hours depending on the governing code, because hydrogen-assisted cracking may not appear for hours after the arc stops. Inspecting hot steel and signing it off is one of the most common and most costly VT errors.

Pre-Weld and Post-Weld VT Checklist

Before the arc: correct base metal and heat number against the drawing; joint type, bevel angle, root gap, and root face per WPS; faying surfaces free of oil, mill scale, paint, moisture, and oxide; fit-up alignment within tolerance with sound, qualified tacks; consumables matching the WPS with current storage and baking records; preheat applied and measured where required.

After cooling: weld size, length, and throat measured rather than estimated; undercut depth and length within acceptance criteria; profile checked for convexity, concavity, overlap, and toe transition; craters, arc strikes, spatter, and gouge marks noted; cracks found visually and confirmed by PT or MT before any repair; distortion and dimensions verified against the drawing.

Done this way, VT stops being the cheap test nobody respects and becomes what it actually is: the highest-yield inspection money you will spend on any weld, and the first line of defence for every method that follows.

Surface and Near-Surface Testing: Liquid Penetrant (PT) and Magnetic Particle (MT)

PT and MT are the workhorses for surface and near-surface discontinuities. Neither replaces volumetric NDT, but both catch the defects that matter most at the skin: toe cracks, crater cracks, undercut, grinding tears, and porosity that has broken through. They are cheap enough to use routinely and sensitive enough to embarrass a weld that looked fine under shop lights. The trick is picking the right one and running it properly.

Liquid Penetrant Testing: Capillary Action in Plain English

PT relies on capillary action. Clean the surface, flood it with low-viscosity penetrant, let the liquid seep into any opening, remove the excess, then apply a developer that draws the trapped penetrant back out. A crack invisible to the eye becomes a bright line against a white background. It finds only surface-breaking discontinuities on non-porous materials. A casting with surface porosity will light up like a constellation and tell you little about the weld.

Pre-cleaning is the whole ballgame. Oil, paint, oxide, moisture, and smeared metal can seal a crack or dilute the penetrant. The sequence is cleaner, penetrant, dwell, removal, developer, inspection. Water-washable penetrants rinse directly; post-emulsifiable systems need an emulsifier before rinsing; solvent-removable systems are wiped clean and suit spot work. Fluorescent penetrants need UV-A light and a darkened area, while visible dye works under white light and is easier in the field. ASME BPVC Section V, Article 6 covers the technique, and dwell times and temperature limits exist because cold penetrant turns sluggish and hot surfaces can flash off the carrier.

AISC notes that PT gives clearly visible indications against a contrasting background, but it is slow and depends on proper cleaning before the developer is applied.

Magnetic Particle Testing: Magnetism With a Direction

MT uses magnetism. Magnetize a ferromagnetic part, apply iron particles, and look for clusters at a flux leakage field. A surface crack or near-surface inclusion disturbs the field, and particles gather along the disturbance. The field must run roughly perpendicular to the defect; a crack parallel to the field may barely register. That one rule explains most of the technique.

Yokes are the field tool for weld toes and small areas. Prods inject current and create a circular field around the contact points. Coils produce a longitudinal field for parts that fit inside. AC fields concentrate near the surface and suit surface-breaking cracks; DC fields penetrate deeper. Wet fluorescent media is the sensitive shop choice, while dry powder is portable and forgiving on rough surfaces. Demagnetize afterward when residual magnetism could affect machining, welding, or arc stability. ASME BPVC Section V, Article 7 covers the method.

MT only works on ferromagnetic material. Austenitic stainless steel, aluminum, copper, and most nickel alloys are out. For aluminum fabrication, our aluminum MIG welding guide covers the welding side; for inspection, PT or another method is the route. Ferromagnetic steels, including the alloy grades in our alloy steel welding electrodes guide, respond well. The same AISC guidance notes that MT is faster than PT for surface-breaking flaws but is preparation-dependent, since surface ridges can trap particles and mimic defects.

AWS B1.10 groups PT and MT among the lower-cost methods compared with RT, UT, and ET, which is why they are often the first line of defense.

PT vs. MT at a Glance

Factor	PT	MT
Materials	Any non-porous solid	Ferromagnetic only
Finds	Surface-breaking flaws	Surface and near-surface flaws
Defect orientation	Any opening to the surface	Best when perpendicular to the field
Speed and cleanup	Slow; chemicals and dwell time	Faster; less cleanup on simple parts
Relative cost	Low	Low
Main false-call risk	Rough surfaces, scratches, poor cleaning	Surface ridges, arc strikes, wrong field direction

Field Workflows That Hold Up

PT, step by step:

1. Clean and dry the surface; confirm it is free of oil, paint, and oxide.
2. Apply penetrant and allow the specified dwell time.
3. Remove excess penetrant using the approved method for that system.
4. Apply a thin, even developer coating.
5. Inspect under the correct lighting within the developer dwell time.
6. Record indications, clean the part, and compare results with the acceptance criteria.

MT, step by step:

1. Clean the surface and remove loose scale or spatter.
2. Verify field direction with a pie gauge or field indicator.
3. Magnetize the area using the approved yoke, prod, or coil technique.
4. Apply wet or dry media while the field is active.
5. Inspect under adequate light and look for clustering, not just dark specks.
6. Demagnetize if required, clean the part, and document the results.

Used together, PT and MT cover most surface-breaking weld defects at modest cost. Used carelessly, they produce confident-looking reports that mean very little.

Radiographic Testing (RT): Seeing Through the Weld, and What It Still Cannot See

Radiographic testing is the only common weld inspection method that hands you a picture of the inside of a joint. A radiation source on one side, a detector on the other, and the weld in between. Because denser material absorbs more radiation, thickness changes, porosity, slag, and tungsten leave a shadow image on film or a digital panel. You are not guessing at what is buried in the metal. You are looking at it.

That picture is seductive, and it is also where most misunderstandings about RT begin. The image is a two-dimensional shadow of a three-dimensional object, and that single fact governs everything RT can and cannot find.

X-Ray Tubes vs. Gamma Isotopes

Industrial radiography uses two kinds of sources. An X-ray tube accelerates electrons into a target to generate radiation, and shutting it off stops the beam instantly. Gamma sources are radioactive isotopes inside a shielded exposure device — iridium-192 and selenium-75 for common wall thicknesses, cobalt-60 for heavy sections — and they emit continuously until the source is cranked back into its shield. X-ray machines offer adjustable energy and no source decay; isotopes offer portability and no electrical supply, which is why they dominate field work on pipelines and tanks. Very thick sections sometimes call for linear accelerators, which behave like very powerful X-ray tubes.

Film, Computed Radiography, and Digital Detector Arrays

Traditional film still has loyal users because it is well understood, cheap per shot, and legally accepted almost everywhere. Computed radiography uses a phosphor imaging plate read by a laser scanner; digital detector arrays convert radiation directly into pixels. Both digital routes cut exposure time, eliminate darkroom chemistry, and store images with metadata for later review. Digital images can be windowed and levelled after the fact, which rescues marginal shots that film would have lost. The trade-off is capital cost, detector care, and a calibration discipline — bad pixel maps and gain corrections — that film never demanded.

Density, Image Quality Indicators, and Geometric Unsharpness

Three technical levers decide whether a defect appears at all. Density is how dark the image is; too light or too dark and small features vanish. Image quality indicators — small plaques containing wires or drilled holes of known size — are placed on the part to prove the technique can actually resolve what it claims. Geometric unsharpness is blur caused by source size and distance: a smaller source, a longer source-to-object distance, and the detector pressed tight against the part all sharpen the image. These are not preferences. AWS B1.10 and ASME BPVC Section V Article 2 spell out the acceptable ranges, and a shot that ignores them is a shot nobody can defend.

The Orientation Problem: Why RT Misses Tight Cracks

Here is the limitation that gets fabricators into trouble. Radiation casts a shadow along its own path, so a discontinuity only shows clearly when its plane roughly lines up with the beam. AWS B1.10 notes that radiography can detect discontinuities whose major planes are parallel with the beam, such as cracks oriented normal to the material surface — and that it usually cannot detect laminations or cracks lying parallel to the plate surface. The AISC engineering FAQ makes the same point: RT finds volumetric inclusions reliably but misses planar separations parallel to the surface. Porosity, slag, and incomplete penetration with an opening show up clearly. A hairline crack running through the thickness of a plate is often invisible, because the beam passes along its length and picks up almost no extra absorption.

Radiation Safety Is Not a Formality

RT is the only weld NDT method that can hurt you from across a room. The industry works to ALARA — as low as reasonably achievable — and ASNT’s radiation safety guidance reduces that to three levers: time, distance, and shielding. Halve the time, double the distance, put steel or concrete between you and the source. In the United States, industrial radiography is licensed by the Nuclear Regulatory Commission under 10 CFR Part 34, with agreement states running parallel programs, and worker exposure limits sit in OSHA 1910.1096. Radiographers carry dosimetry and, for gamma work, an IRRSP card that requires 40 hours of formal radiation safety instruction plus 320 hours of gamma experience — 160 hours if they only shoot X-ray.

What RT Costs You in Time and Scheduling

All of that safety apparatus has commercial consequences. A radiograph means barricades, warning signs, an exclusion zone, a cleared area, and usually a second person standing watch. On a live plant, that can mean a permit, a shutdown window, and coordination with everyone else who needs the same space. Shots are also slower than they look: setup, positioning, exposure, processing, and review. When a specification calls for 100% RT on a heavy wall, the schedule impact is real. That is one reason digital radiography keeps gaining ground — and one reason phased array ultrasonic testing has quietly replaced RT on many thick-section and pipeline jobs, especially where access, cost, and crack detection all point the same direction. Aerospace work, where wall thicknesses are thin and records must survive decades, remains a stronghold for film and digital RT alike; the same documentation culture shows up across aerospace welding programs generally. Aluminum fabrications are another natural fit, since castings and thin aluminum joints respond well to radiographic inspection, though surface prep still matters as much as it does in aluminum MIG welding.

The honest summary: RT is unmatched for volumetric defects and gives you a permanent visual record, but it demands geometry, technique, and paperwork discipline to be worth the radiation dose and the schedule disruption. Ask for it when porosity, slag, or internal inclusions are the risk. Do not ask for it because it sounds more thorough than a penetrant test.

Ultrasonic Testing (UT) and Phased Array (PAUT): The Method That Finds Planar Flaws

Radiography gets the reputation; ultrasonics does the hard jobs. When a weld is thick, when the flaw you fear is a crack rather than a pore, or when you need to know how tall a defect is rather than merely that it exists, UT is usually the right answer. It is also the most operator-dependent method in the NDT toolkit.

How Pulse-Echo UT Actually Works

A piezoelectric transducer fires a short pulse of high-frequency sound, typically 2 to 10 MHz, into the steel. Part of that energy reflects off any boundary and returns; the machine times the round trip and draws it as a spike. Couplant — gel, oil, glycerin — exists only to bridge the air gap between probe and surface, because sound will not cross air usefully.

For welds, technicians rarely shoot straight down. Angle-beam probes (45°, 60°, 70° in common nomenclature) refract into a shear wave and bounce zig-zag through the thickness, reaching the fusion faces where lack-of-fusion and sidewall cracks live. Before scanning, the technician calibrates on a reference block with known reflectors and builds a distance-amplitude correction (DAC) curve: the expected echo height from a reference reflector at every depth. Any indication rising above a set percentage of that curve is recorded. Without a valid DAC, the numbers on screen mean nothing.

Crack or Slag? Reading Echo Behaviour

A slag line is volumetric and rounded, so its echo behaves predictably as the probe is rotated and walked. A crack is planar and sharp: it returns a strong, brittle echo at one specific angle and drops away fast when you move off it, and you can often pick up tip-diffraction signals from its upper and lower extremities.

Orientation still rules. AWS B1.10 is careful here: ultrasonic examination can detect cracks oriented in any direction, provided the beam is aimed essentially perpendicular to the crack’s major axis. The AISC’s own guidance on welding inspection puts the trade-off plainly — UT avoids radiation hazards and finds planar defects parallel to the surface, but may miss planar flaws perpendicular to it.

Phased Array: Steering the Beam Electronically

A conventional probe holds one crystal at one angle. A phased array probe holds dozens, fired in sequence with tiny time delays. Staggering those delays steers, focuses, and sweeps the beam without moving anything. ASNT’s overview of phased array UT notes that this electronic control improves probability of detection, and that electronic scanning typically covers components an order of magnitude faster than a single-probe mechanical system.

Two display types matter. The S-scan is a sectorial picture — the beam swept through a range of angles, producing a cross-sectional image of the weld. The E-scan is a line of focal laws fired along the array for a fast linear sweep. On encoded scans, a position encoder tracks the probe so data can be plotted against real distance, stored, and re-examined later. That record is why PAUT has displaced radiography on much thick-wall and pipeline work: an image, a data file, and a permanent archive with no radiation exclusion zone.

Governing documents are worth knowing. ASTM E2700 covers contact phased-array testing of welds from 9 to 200 mm (0.375 to 8 in). ASME BPVC Section V Article 4 sets general ultrasonic requirements, and ASME Code Case 2235 permits advanced ultrasonic techniques, including automated UT, in defined circumstances.

TOFD as a Sizing Companion

Time-of-flight diffraction uses two probes straddling the weld, one transmitting and one receiving. Instead of echo height it listens for weak diffracted signals from the top and bottom tips of a flaw, giving remarkably accurate through-wall height sizing. Its weakness is a dead zone near the surface and back wall, so in practice TOFD runs alongside pulse-echo or PAUT rather than replacing it.

Why UT Is the Most Operator-Dependent Method

The AISC does not soften this point: ultrasonic testing takes skill, and results depend heavily on the technician. A poor radiograph is still a poor radiograph anyone can review. A poor UT scan may be a clean screen produced by someone using the wrong angle, the wrong gain, or the wrong couplant. That is why the written procedure, the joint-specific scan plan, and genuine Level II and Level III competence carry more weight here than anywhere else. On safety-critical work — the kind of standard you see in aerospace TIG welding — the scan plan is often reviewed before a probe touches metal.

Cost and Access Trade-offs

UT carries no radiation burden, needs only one side of the joint, and gets cheaper per joint as wall thickness climbs, because RT exposure times and film costs rise with thickness while UT does not. What it adds is surface preparation, couplant cleanup, and slower manual scanning on complex geometry. Where the code allows substitution it is a code question, not a preference: ASME Section VIII Division 1 UW-51 permits ultrasonic examination in lieu of radiography where the thinner joined material is at least 0.25 in.

Beyond the Big Five: Eddy Current, Leak Testing, Acoustic Emission, and Emerging Tools

Most welding inspections never get past visual, penetrant, magnetic particle, radiography, and ultrasonics. That is understandable. Those five methods cover the overwhelming majority of shop and field work. But they are not the whole toolbox. AWS B1.10 lists seven standard NDE methods for welds, and two of the extras, electromagnetic testing and leak testing, solve problems the big five cannot. A few more specialized techniques are worth knowing by name, even if you only meet them once or twice in a career.

Eddy Current Testing: For Conductive Surfaces

Eddy current testing (ET) works on a simple principle. An alternating current in a coil creates a changing magnetic field, which induces small circulating currents in a conductive test object. Cracks, changes in conductivity, or variations in lift-off disturb those currents, and the coil picks up the change. The catch is right there in the physics: the material must conduct electricity. Carbon steel, stainless steel, aluminum, copper alloys, and titanium are fair game. Plastics and most composites are not.

ET is strongest on surface and near-surface discontinuities, especially in thin material, tube walls, and weld overlays. It is fast, needs no couplant, and can inspect coated or painted surfaces within limits. It is also the backbone of heat exchanger tube inspection, where a probe travels down the inside of thousands of tubes. For everyday structural welding, it is a specialized service rather than a routine call. If you work with aluminum, remember that surface condition matters enormously; poor cleaning can create misleading signals. Our guide to aluminum boat welding covers why hulls and marine structures demand clean, well-prepped surfaces before any inspection.

Leak Testing: Proving a Vessel Holds

Leak testing (LT) answers a different question. It does not tell you whether a weld is strong. It tells you whether the assembly leaks, and sometimes where and how much. The common variants are bubble immersion, pressure decay, helium mass spectrometry, and halo-halide detection. Bubble testing is cheap and visual. Pressure decay is simple but sensitive to temperature changes. Helium mass spectrometry is the precision instrument, capable of finding extremely small leaks, and it is standard practice for vacuum systems, cryogenic equipment, and high-purity piping.

Leak testing is usually a system-level test, not a weld-level test. That distinction matters. A leak can come from a weld, a gasket, a fitting, or a valve stem. The test proves the assembly, and then the detective work begins. It is also sensitive to test conditions, so the written procedure needs to specify the gas, pressure, stabilization time, and acceptance limit.

Acoustic Emission and Magnetic Flux Leakage: Listening and Mapping

Acoustic emission (AE) is passive. Instead of sending energy into the part, it listens for the stress waves released when a flaw grows or a crack tip advances under load. That makes it useful for monitoring tanks and pressure vessels during a proof test or during service. ASME BPVC Section V Article 12 covers acoustic emission examination of pressure vessels. AE is not a substitute for a full volumetric inspection. It tells you that something is active and roughly where, then other methods confirm the finding.

Magnetic flux leakage (MFL) is the workhorse of storage tank floors and pipeline inspection. A magnetizer saturates the steel, and sensors detect flux that leaks out at corrosion pits, cracks, or wall loss. Tank floor scanning and in-line pipeline pigs are mature, highly specialized applications. You hire a service, not a lone technician with a yoke.

Emerging Tools: More Data, Fewer Blind Spots

Encoded phased array scanning is already changing field UT. Because the probe position is recorded, the data can be stored and reprocessed later, something conventional manual UT cannot provide. That is a major advantage when a joint is buried, coated, or too expensive to revisit. Automated and robotic UT crawlers extend the same idea to tanks and long seams. Laser ultrasonics, thermography, and shearography offer non-contact or full-field screening, mostly in aerospace, composites, and research settings. DICONDE formats are making inspection images easier to archive and audit, while machine-assisted indication flagging helps technicians sort through large data sets. None of these replace a qualified Level II with a written procedure. They change how much data that person can handle.

Before you book a specialized service, ask the vendor four questions. Which procedure and revision will you follow? What is the technician’s certification level and who employs them? What acceptance criteria apply, and who wrote them? What deliverable will I receive, and can I get the raw data? If the answers are vague, keep shopping. The exotic method is not the hard part. The paperwork and competence behind it are.

Choosing the Right NDT Method: A Practical Decision Framework

Most method arguments start in the wrong place. Someone says “let’s UT it instead of RT,” and the room divides into two camps before anyone has opened the drawing. The right question is never which method is best. It is which method satisfies the governing document, finds the discontinuity you actually expect, and fits the geometry, surface, and schedule you actually have.

Start with the code. It is not a starting suggestion; it is the ceiling and the floor at the same time. In structural steel, AWS B1.10 frames the selection problem clearly: the choice turns on method advantages and limitations, the anticipated discontinuity type and size, the acceptance standards in play, and cost. That is a four-question filter, and it works on any job.

Step 1: Read What the Code Already Mandates

Before selecting anything, find out what you are not allowed to choose. AWS D1.1/D1.1M Clause 8 provides the inspection methods and acceptance criteria for structural steel welding, so on a D1.1 job the method list is largely settled by clause and joint type.

AISC Specification Chapter N goes a step further and requires a QA/QC plan that names inspection methods and their frequencies. Two distinctions from AISC are worth memorizing:

• When the Structural Engineer of Record specifies RT or UT, it implies total internal soundness matters and the weld must meet the acceptance standard.
• When VT, MT, or PT are specified, internal soundness is expected to come from adherence to AWS D1.1, because those methods only see the surface or near-surface.

AISC also warns that RT and UT have practical limits tied to joint geometry and material thickness. That warning is the whole reason a single method rarely covers a whole project.

Step 2: Work Through the Selection Variables

Once the mandatory methods are locked in, the remaining choices follow a predictable sequence:

1. Material. Ferromagnetic only for MT. Conductive for ET. Austenitic stainless and aluminum rule MT out entirely, which pushes surface work toward PT.
2. Thickness. Thin sheet favors RT; heavy wall favors UT and PAUT, where penetration and depth sizing work in your favor.
3. Joint geometry and access. A backing bar, a double-wall configuration, or a nozzle-to-shell intersection can make a method blind regardless of how good it is.
4. Expected discontinuity. Volumetric (porosity, slag) reads well on RT. Planar and tight (lack of fusion, cracks) needs angle-beam UT or PAUT.
5. Surface condition. PT and MT need clean, paint-free, reasonably smooth surfaces. Grinding to 125 µin. is common for PT.
6. Schedule and radiation. RT brings exclusion zones, dosimetry, and often night shifts. A licensed radiation program is not a same-day step.

Comparison Table: What Each Method Delivers

Method	Finds	Misses	Material Limits	Relative Cost	Typical Use
VT	Undercut, mismatch, porosity at surface, cracks, fit-up	Everything internal	None	Lowest	Every weld, all phases
PT	Surface-breaking cracks, laps, porosity	Subsurface and tight closed defects	Nonporous surfaces only	Low	Stainless and aluminum welds
MT	Surface and near-surface planar flaws	Deep internal flaws	Ferromagnetic only	Low to moderate	Structural steel, fillets
RT	Volumetric porosity, slag, tungsten, burn-through	Tight planar cracks parallel to the beam	Thickness and geometry limits	Moderate to high	Pipe, thin plate, code-mandated CJP
UT/PAUT	Planar flaws: lack of fusion, cracks, deep slag	Some near-surface and complex geometry zones	Requires access and couplant	Moderate	Heavy wall, thick-section CJP, pipeline
ET	Surface and near-surface in conductive metals	Deep internal flaws	Conductive materials only	Moderate	Tubes, cladding, overlay
LT	Through-leaks and pressure loss	Any flaw that does not leak	Needs a pressure boundary	Varies widely	Tanks, vessels, sealed components

Step 3: Set the Extent and Frequency

Method selection and inspection extent are separate decisions, and confusing them causes most disputes. Extent is expressed as spot versus 100%, or as a sampling plan tied to joint type, welder, or shift. On structural work, Chapter N of the AISC Specification sets the QA/QC framework, while AWS D1.1 Clause 8 carries the method and acceptance detail.

Bridge work shows how prescriptive this gets. FHWA research notes that AWS D1.5 requires RT for complete joint penetration welds subject to tensile or reversal stresses, while either RT or UT may be used where the stress is compressive or shear. Same weld detail, different NDT, purely because of the stress state.

Three Worked Scenarios

Structural steel moment connection. Demand is CJP groove welds at the flanges, ferromagnetic material, thickness typically 1 inch and up. VT on every pass, then UT or PAUT on the CJP welds per the engineer’s direction. RT is usually a poor fit here because the geometry and thickness stack against it. Magnetic particle is useful for the access-hole and backing areas. If the base metal is a quenched-and-tempered grade, remember that surface prep and inspection timing matter as much as the method.

Thin stainless process piping. Austenitic stainless kills MT, so surface work falls to PT with a fluorescent penetrant and careful cleaning. For the root and fill passes on thin wall, RT remains a strong volumetric option and is often what the piping code and client specification expect. UT on thin austenitic wall is difficult because of grain scatter and geometry, which is exactly the kind of limitation AISC flags. Weld cleaning discipline decides whether PT gives you a real answer or a false call, and the same prep logic appears in guides like aluminum cleaner for welding when surface condition governs the result.

Heavy-wall pressure vessel. Thickness here is the deciding factor. PAUT with encoded scanning covers the full volume, records the data for later review, and pairs well with TOFD for sizing. RT becomes impractical as wall thickness climbs, both for exposure time and for crack orientation. Add leak testing at hydro or pneumatic stage, and MT or PT on nozzle and attachment welds depending on material.

Document the reasoning. A one-page selection note listing the governing clause, the mandatory method, the expected discontinuity, and the extent you chose will end more arguments than any amount of experience. When the material itself is unfamiliar, background reading helps too; aluminium work, for instance, brings its own inspection quirks covered in aluminum MIG welding resources.

People, Procedures, and Paperwork: What Separates a Real Inspection from a Rubber Stamp

Two ultrasonic reports can look identical on the page and mean completely different things. One came from a certified Level II technician working to a written procedure, on a calibrated machine with records on file, signed by someone accountable. The other came from a contractor who “has been doing this for twenty years,” ran a quick scan, and typed a paragraph. Both cost money. Only one is evidence.

This is the human and documentary side of NDT, and it is where most inspection failures actually start. Not in the physics.

The Three Levels, and What They Actually Mean

NDT personnel are qualified at three basic levels. The structure is standard across the industry:

• Level I: Performs specific calibrations and tests under written instruction, then reports results. Cannot interpret independently in most schemes.
• Level II: Sets up and calibrates equipment, performs and interprets the test per the procedure, and reports. This is the person doing the work on the floor.
• Level III: Develops procedures, establishes techniques, interprets codes and standards, and owns the overall NDT program. Often the name on the procedure nobody reads.

One detail worth remembering: levels are method-specific. A Level II in ultrasonic testing is not a Level II in radiography, and a card that lists a level without a method is only half a credential.

Qualification vs. Certification: Not the Same Word

ASNT draws a distinction that gets blurred constantly. Qualification is the process: training, experience, and examination that demonstrate the person can do the job. Certification is the written testimony that the individual has been properly qualified. You cannot have the second without the first, but plenty of paperwork pretends otherwise.

Two documents govern how employers build that system. ASNT SNT-TC-1A provides a framework for employer-written qualification and certification programs, with recommended education, experience, and training hours by method. It is a guideline the employer tailors. ANSI/ASNT CP-189 is different: it sets minimum requirements that may not be reduced. If your written practice quietly trims CP-189 minimums, it is no longer CP-189 compliant.

Internationally, ISO 9712 takes a different route, certifying individuals through an independent body rather than through the employer’s own program. The employer still has to authorize the person to operate on its equipment, under its procedures, on its work. Certification travels. Authorization does not.

The Written Procedure Is the Rulebook

A procedure is not a formality. It defines the method, the technique, the equipment, the search unit, the scan pattern, the coverage required, calibration intervals, and the acceptance criteria. Those acceptance criteria must trace back to the code the job is being built to, whether that is ASME Section V for the examination method or a structural welding code for the accept/reject line. A procedure that says “acceptance per customer specification” and stops there is not a procedure. It is a placeholder.

Calibration and Traceability

Equipment drifts. That is why calibration is checked at defined intervals, at the start of a shift, and again whenever a transducer or cable is swapped. Reference blocks and calibration certificates should trace to a national standard. If nobody can produce a calibration date or a serial number, the readings are anecdotes.

What a Real Report Contains

A defensible report is identifiable and traceable. At minimum it carries:

• Project, drawing, and specific weld or component ID
• Method, technique, and the procedure number with revision
• Equipment and instrument serial numbers, plus transducer details
• Calibration references and the code acceptance criteria applied
• Results with location, size, and classification of indications
• Disposition: accept, reject, or repair
• Date, technician name, level, certification reference, and signature
• Level III review or authorized inspector sign-off where the code requires it

Red Flags of a Rubber Stamp

• No procedure number, or a revision that predates the job
• “Passed NDT” with no method, coverage, or instrument named
• “100% UT” claimed without a coverage plan or scan map
• Certification from a quiz, with no practical examination behind it
• The identical report template for every joint on every project

The paper trail is the product. If it was not written down, traceable, and signed, it did not happen.

Cost, Mistakes, Repairs, and a Practical Playbook for Your Next Weld

Most of the money in NDT is not spent on the testing itself. It is spent on the consequences of testing late, testing wrong, or testing without a document that says what “pass” means. The commercial side and the technical side are the same conversation.

How NDT Is Actually Priced

Nondestructive testing is priced like a service call, not a product on a shelf. A typical quote breaks into mobilization and travel, a minimum call-out that is often two to four hours even when the job takes twenty minutes, a per-foot or per-joint rate, and consumables such as film, chemistry, couplant, and the amortized cost of calibration blocks. Then come the charges nobody budgets for: night-shift premiums, report and paperwork time, and re-shoots or re-scans when an image is rejected for density, geometry, or technique error.

That is why the cheapest quote is often the most expensive line in the project. A low per-joint rate from a crew running without a written technique sheet, with expired calibration stickers, or with a Level I signing work that needs a Level II can produce a report your engineer or client will not accept. You pay twice, and the second payment includes the schedule slip.

The Mistakes That Cost the Most

• Inspecting too early. Hydrogen-assisted cracking can appear hours or even days after welding, which is why many specifications impose a hold period before final inspection of crack-susceptible steels and thick sections. Testing at hour two gives you a clean report on a weld that cracks tonight.
• Painting before visual testing. Coating hides undercut, spatter, arc strikes, and toe cracks. Once primer is on, VT becomes a coating inspection.
• Choosing the wrong method for the defect orientation. PT and MT only see the surface. RT struggles with tight planar cracks lying parallel to the plate surface. That is a UT job.
• Skipping the written procedure. Two technicians can test the same joint and reach different conclusions, and neither result is defensible.
• Expired calibration. A single out-of-date sticker can invalidate an entire day of results.
• No acceptance criteria in the contract. AWS B1.10 is clear that discontinuities may be acceptable, provided their size and distribution fall within specified limits. The keyword is specified. If nobody wrote the limits down, nobody can accept or reject against them.
• Arc strikes left in place. They are stress concentrators and hardened zones, and many criteria reject them outright regardless of size.
• Grinding out indications without engineering approval. Chasing a slag line with a burr can quietly take wall thickness below minimum. Removal is an engineering decision, not a shop-floor one. Gouging is its own process with its own rules, so read this guide to air arc welding and carbon arc gouging before you start.
• Treating a repair as a fresh start. A repair weld is a new weld. It generally needs a qualified welding procedure and qualified welders, with ASME BPVC Section IX governing that qualification, plus re-examination by the same method, including the repaired area and adjacent material.

One contract detail catches people every year. AISC notes that rework required to correct internal discontinuities found by RT or UT is part of the contract requirement when those methods are specified. Because VT, MT, and PT do not describe internal condition, owner-required rework of internal discontinuities found by other means is treated as a change in contract requirements. Translation: who pays for the gouge and re-weld depends on what the drawings asked for.

Playbook: From Drawing Review to Final Report

1. Read the drawing and the governing code. Identify the method, the extent, and the acceptance criteria.
2. Confirm those criteria are written into the contract documents, not assumed.
3. Verify joint preparation and fit-up with pre-weld VT before anybody strikes an arc.
4. Confirm the WPS is qualified and the welders are qualified for it.
5. Schedule NDT around cooling and any required hold time, not around the shipping date.
6. Perform VT first, after cleaning and before coating.
7. Run the specified method to the written technique.
8. Collect the report, calibration records, and technician certifications in one package and file them with the weld map.

Repair and Re-Test Workflow

1. Characterise the indication: size, location, orientation, depth.
2. Get engineering disposition: accept, repair, or monitor.
3. Remove material by approved method, then re-test the excavation to confirm the flaw is gone.
4. Re-weld under a qualified procedure with a qualified welder.
5. Re-examine the repair plus adjacent base metal by the original method.
6. Record the repair on the weld map so the next inspector knows the history.

A Short Career Note

NDT is one of the few trades where the entry barrier is knowledge rather than capital, and the ceiling is high. Certification routes through ASNT in the US and ISO 9712 internationally, and radiation safety credentials are a separate track. If you are weighing a move from the torch to the transducer, an aluminium welding course is a useful complement, because knowing how a weld is actually made separates a technician who reads indications from one who understands them.

The Bottom Line: NDT Is Design Feedback

It is tempting to treat inspection as policing, a gate that either passes your work or punishes it. That framing costs money. A crack found in the shop is a process adjustment: preheat, interpass temperature, fit-up, cleaning, technique. The same crack found in service is an outage, a cut-out, and possibly an incident report. NDT is the only honest feedback loop a fabricator gets on work that will otherwise be judged by time. Treat the report as data about your process, and the cost conversation changes entirely.

Frequently Asked Questions

What is an NDT test for welding in simple terms?

It is any examination that checks a weld’s integrity without cutting the part apart. Common methods include visual testing, dye penetrant, magnetic particle, radiography, ultrasonic testing, and phased array. The method chosen depends on the material, thickness, joint geometry, and the type of flaw you expect to find.

Which NDT method is best for welding?

There is no single best method. Visual testing catches surface and fit-up problems cheaply, penetrant and magnetic particle find surface-breaking flaws, radiography excels at volumetric flaws like porosity and slag, and ultrasonic testing is best for planar flaws such as lack of fusion and cracks. Most codes combine methods rather than relying on one.

Can NDT guarantee a weld will never fail?

No. Every method has a probability of detection and blind spots, and results depend on technique, procedure, and technician skill. NDT reduces risk by finding discontinuities above a defined acceptance threshold; it does not certify infinite life. Fitness-for-service and fracture mechanics assessments exist precisely because inspection is statistical.

How soon after welding can a weld be tested?

Surface methods can usually be applied once the weld has cooled enough to handle and the surface is clean and dry. For crack-susceptible steels and thick sections, many specifications require a deliberate delay, often 24 to 48 hours, so hydrogen-assisted cracks have time to appear before final examination.

Is radiographic testing dangerous?

Industrial radiography uses ionizing radiation and is regulated. With proper licensing, barriers, exclusion zones, dosimetry, and trained personnel following ALARA principles, it is a controlled industrial activity. The real risks come from unlicensed operation, uncollimated sources, and people entering an uncontrolled area.

Do I need a certified inspector for weld NDT?

In most code work, yes. Personnel are typically qualified to NDT Level I, II, or III under ASNT SNT-TC-1A, ANSI/ASNT CP-189, or ISO 9712, and radiographers also need radiation safety certification such as ASNT IRRSP plus the employer’s license. The employer certifies the individual, and the certification should be available for review.

How much does weld NDT cost?

Costs vary widely by method, access, and location. Visual testing is cheapest, penetrant and magnetic particle are mid-range, and radiography and ultrasonic testing cost more due to equipment, personnel, and safety requirements. Mobilization and minimum call-out charges often dominate small jobs, so bundling inspections is usually cheaper than repeated visits.

Why did my weld pass visual inspection but fail ultrasonic testing?

Visual testing only sees the surface. Lack of fusion, incomplete penetration, and internal slag can sit entirely beneath a sound-looking cap. That is not a contradiction; it is exactly why volumetric methods are specified for critical joints. It also means a good-looking bead is not evidence of root fusion.

Can a failed weld be repaired and re-tested?

Yes, but a repair is a controlled operation, not a patch job. The indication is typically excavated to sound metal, the repair is welded under a qualified procedure, and the area is re-examined with the same method plus a check of adjacent material. Repeated repair cycles on the same spot can degrade properties and may need engineering approval.

What is the difference between UT and PAUT?

Conventional UT uses a single-element probe at one angle per scan. Phased array UT uses many elements with electronic time delays to steer, sweep, and focus the beam, producing cross-sectional images and enabling encoded data storage. PAUT is generally faster and more repeatable, but it still depends on a qualified procedure and a competent operator.