Take two butt welds in 20 mm carbon steel, single-V prep with a 30° bevel, each holding a flaw about ten millimeters long. The first is a line of porosity in the fill, the second a lack of sidewall fusion lying along the bevel face. Radiography shows the first at a glance and can leave the second almost invisible. Ultrasound does the reverse: a 60° angle probe meets the bevel face nearly head-on and gets a clean echo, while scattered pores send back weak signals that are hard to characterize.
So whoever writes a weld inspection plan picks the method first on the flaw the designer worries about, then on wall thickness, and only after that on access, records and radiation safety. It's also the order the weld NDT standards follow.
Density versus acoustic impedance: what each method records
Radiography measures how much radiation gets through the part along each line of the beam. A pore or a slag inclusion removes or adds material along that line and shows up as a darker or lighter spot on film or panel. Contrast depends on the flaw's volume projected along the beam, which is why porosity, gas cavities, inclusions and open lack of penetration are the method's home ground. A tight crack behaves differently. When its plane lines up with the beam, radiation travels a few millimeters of empty gap and the crack shows. When it's tilted, the gap crossed shrinks to its opening, perhaps a few hundredths of a millimeter, and the signal sinks into noise. The UK Health and Safety Executive's guidance on industrial radiography (2008) puts it plainly. Volumetric flaws are detected more readily than planar ones, crack detection depends on size, morphology, orientation and opening, and the beam should be aligned as closely as possible with the flaw plane, with multiple shots when orientation is unknown. The same guidance flags when other methods should be considered.
Ultrasound works the other way around. A pulse of a few megahertz enters the metal and comes back wherever it meets a jump in acoustic impedance, such as a steel-to-air interface. A broad, flat surface like an unfused bevel face is an excellent reflector when the beam strikes it close to perpendicular, while a spherical pore scatters energy in every direction and returns little. In exchange, time of flight tells you how deep the indication sits and, with the right technique, how tall it is, which a single radiograph, a flat projection of the part, can't provide.
Thickness moves the boundary: what ISO 17635 says
The standard linking material, thickness and method is ISO 17635, the general rules for testing welds in metallic materials. In the 2016 edition, for butt joints in ferritic steel, the table for internal imperfections pointed to radiography up to 8 mm, with ultrasound allowed only with limitations. Between 8 and 40 mm the two methods stood on equal terms, while above 40 mm ultrasound took over, with radiography allowed only with limitations. For austenitic steels radiography remained the reference at all thicknesses, with ultrasound referred to ISO 22825. The 2025 edition now in force changed that very table, added phased array for thin-walled components and the TFM technique, and brought back an annex with a selection flowchart. For a new specification, read the table in the 2025 text.
The same standard translates weld quality level into technique and acceptance criterion for each method. Under the 2016 edition, for ISO 5817 quality level B, digital radiography called for class B and acceptance level 1 of ISO 10675-1. Manual ultrasound called for at least testing level B of ISO 17640 and acceptance level 2 of ISO 11666. For level D, manual ultrasound wasn't recommended. The detail matters, because the same joint can be accepted by one method and rejected by the other: each sees different imperfections and measures them differently. Radiographic acceptance is covered in our piece on who decides acceptance levels.
How much they really find: measured probability of detection
The most useful public figures come from trials on welds with known flaws, and they should be read with their sample attached. HSE's 2008 radiography guidance reports the Dutch NIL study on welded plates up to 15 mm. With a single exposure, X-ray or gamma, probability of detection stopped at 60-70%, rising to 95% with two exposures aligned with the fusion faces. HSE also cites the 1987 NORDTEST trials, over 3,000 interpretations of more than 700 flaws, with an average probability around 70% at the highest sensitivity. No detailed conclusions can be drawn from these data, the guidance warns, but the point about alignment is clear: the second trial worked because the beam was aligned with the lack-of-fusion flaws.
On the ultrasonic side, the best-documented limit is manual sizing. The UK PANI program put a 25 mm double-V joint with an 8 mm high root crack in front of operators, who sized it anywhere from 2 to 10 mm, and the program's second phase showed a similar spread. HSE's guidance on sizing errors (2008) puts the standard deviation of random error for maximum-amplitude sizing at about 3 mm at a 20 mm beam path, rising to 7 mm at 300 mm. It explains why ISO 17640 ties its testing levels to different probabilities of detection, and why recorded ultrasonic inspection, phased array and TOFD, has gained ground in specifications. Georgiou, writing in Insight (2007), adds that a probability-of-detection curve holds for the method, sample and procedure it was measured on, and can't be carried blindly from pipe to plate.
Manual UT, phased array and TOFD don't produce the same report
Writing "ultrasonic testing" into a specification isn't enough, because the techniques have different standards, minimum thicknesses and criteria. Manual testing follows ISO 17640:2018 for joints from 8 mm upward in low-attenuation materials, with three testing levels A, B and C and a fourth, D, reserved for special applications and defined by specification. Acceptance criteria are in ISO 11666:2018, covering 8 to 100 mm. Automated phased array follows ISO 13588:2019 from 6 mm, assessing by amplitude and length or by height and length, with acceptance levels in ISO 19285:2017. TOFD, which measures flaw height from diffraction at the flaw tips, follows ISO 10863:2020 from 6 mm, with acceptance under ISO 15626:2018 between 6 and 300 mm. The TFM technique has its own standard, ISO 23864:2021, from 3.2 mm.
For the buyer, what counts most is the difference between an inspection that leaves an archived image and one that leaves a written record. Phased array and TOFD record the scan with an encoder, so the data can be reread, compared over time and checked by a third party, just like a radiograph. ISO lists several of these standards as under revision, so specifications should cite them with their edition. Instruments and setups are discussed in our phased array guide for welds and in the one on choosing a PAUT and TOFD instrument.
Choosing a method for a new specification?
Send us joints, wall thicknesses, material and governing standard: we'll tell you whether radiography, phased array or TOFD fits, with a field demonstration on your part and a tailored quote within 24 hours.
Email us Message us on WhatsAppWeld inspection, case by case
| Situation | Starting method | Reason |
|---|---|---|
| Ferritic steel up to 8 mm | Radiography | Ultrasound was allowed only with limitations in ISO 17635:2016 |
| Ferritic 8-40 mm, porosity and inclusions expected | Radiography or ultrasound | Both allowed; radiography documents volumetric flaws better |
| Ferritic above 40 mm | UT, PAUT or TOFD | Radiography only with limitations; energy and exposure time climb |
| Risk of cracks and lack of fusion in thick joints | PAUT or TOFD | Planar reflectors suit sound, height can be measured |
| Austenitic steels, nickel alloys, dissimilar joints | Radiography | Ultrasound needs a dedicated procedure under ISO 22825 |
| Access from one side, busy work area | UT or PAUT | Radiography needs source and detector on opposite sides and a controlled area |
Treat the table as a starting point, since the product code or the customer may impose a different method. Where the choice is open, the criterion is the one from the opening example: if the worrying flaw is volumetric and walls are thin, radiography is the safer choice; if it's planar and walls are thick, phased array with TOFD finds more and sizes better.
Where radiography remains the sounder choice
On thin walls ultrasound loses ground fast. Below a few millimeters the flaw echo blends with surface and root echoes, the zone close to the probe takes up a large share of the thickness, and sizing becomes uncertain. Radiography, on the same wall, works at low voltage with high contrast. The same goes for castings, where complex geometry multiplies shape reflections and the typical flaw, scattered porosity, is exactly what sound returns worst.
Then there are coarse-grained materials. In austenitic welds, nickel-alloy welds and dissimilar joints, the columnar structure of the weld metal bends and attenuates the beam, creates structural echoes and makes propagation direction-dependent. Ultrasound is still possible, but with a procedure developed and validated case by case under ISO 22825:2017, which sets no acceptance levels and calls for test blocks representative of the real joint. On a handful of parts, that preparation cost makes radiography simpler. Finally, the reader of the report matters. A radiographic image can be interpreted without knowing the technique behind it, and many product specifications in piping and pressure vessels still call for it by name, as the case studies on welds and castings show.
Who decides the method in the inspection plan
The method is set first by whoever owns the product code or the customer specification. The construction code for the vessel, pipeline or structure states how much of the welding must be inspected and by which permitted methods. Within that boundary the choice passes to the welding coordinator, who writes it into the inspection plan with the technique, the class or testing level and the acceptance criterion, and gets it approved by the client. When X-rays make it into the plan, what's left is deciding how to organize weld radiography, in-house or through an outside service. The inspection is then carried out by personnel certified to ISO 9712:2021 for that method, since a level 2 in radiography doesn't sign an ultrasonic report.
The job site enters the calculation too. Radiography needs a controlled area and, often, a stop to nearby work during exposures, plus radiation protection obligations. Ultrasound, which uses no ionizing radiation, works alongside active crews. When the two methods are combined, the most common answer in demanding specifications, radiography documents volumetric flaws and ultrasound covers planar ones and measures their height. An overview of ultrasonic instruments and techniques is on our industrial ultrasonic testing page.
What to send us for an assessment
To recommend the right method and equipment for inspecting your welds, we need these data:
- material, joint type, wall thicknesses and diameters, with the weld prep drawing;
- product code or customer specification, ISO 5817 quality level;
- the flaws of greatest concern: porosity, inclusions, lack of fusion, cracks;
- access to the joint: one side or both, shop or field;
- volumes: joints per day or per job, and available time;
- method used today and certified personnel available.
You can send them by email to info@dipitech.com or on WhatsApp, with drawings and specifications.
The order of the questions matters more than any single answer. Start from the flaw the designer fears and the wall thickness, check what the product code and the current ISO 17635 allow, and only then look at cost, time and radiation safety. If your case sits in a gray zone, the quickest way to settle it is a field demonstration on your joint, with a tailored quote within 24 hours and technical support in Italy. On the radiographic side, sources and detectors are presented on our radiographic weld inspection page.
Frequently asked questions about radiography and ultrasonic weld inspection
Is radiography or ultrasonic testing better for weld inspection?
It depends on the expected flaw and wall thickness. Radiography detects volumetric flaws such as porosity and inclusions better and works well on thin walls; ultrasound, especially phased array and TOFD, finds planar flaws such as cracks and lack of fusion better, measures their height and prevails on thick walls.
Which standard indicates the method by thickness?
ISO 17635, now in its 2025 edition, which sets general rules for testing welds in metallic materials. In the 2016 edition, for ferritic steel, radiography was the reference up to 8 mm, the two methods were on equal terms between 8 and 40 mm and ultrasound prevailed above that; the table changed in 2025 and should be read in the current text.
Can ultrasonic testing replace radiography on welds?
Often, yes, if the product code or customer allows it and the thickness falls within the ultrasonic standards' scope: from 8 mm for manual testing under ISO 17640:2018 and from 6 mm for automated phased array and TOFD. On thin walls and austenitic joints, radiography usually remains simpler and more reliable.
What does TOFD find that radiography doesn't?
TOFD measures flaw height through the wall, deriving it from diffraction at the flaw tips; because it relies on diffraction rather than reflection, it depends less on flaw orientation. Radiography returns a flat projection of the part, with no depth information from a single exposure.
What is radiography's probability of detection for cracks?
It depends heavily on beam alignment. In the Dutch NIL study reported by HSE's 2008 guidance, on welded plates up to 15 mm a single exposure gave probabilities of detection between 60 and 70%, while two exposures aligned with the fusion faces reached 95%.
How much does weld inspection by radiography or ultrasound cost?
It depends on method, number and thickness of joints, access, governing standard and equipment, so a generic price would be misleading. DIPITECH prepares a tailored quote within 24 hours for the right equipment and can arrange a field demonstration on your joint.