2D radiography answers the question "is there a flaw?". Industrial computed tomography (CT) answers questions that are far more expensive to get wrong: where exactly is the flaw, how large is it in three dimensions, how porous is the critical zone, does the internal geometry match the CAD? This guide explains how industrial CT works, which parameters really determine its performance, and how to decide whether your application calls for CT, 2D radiography or ultrasonic testing.
The physics: from hundreds of projections to a 3D volume
A radiograph is a projection: the X-ray beam passes through the part and the image records the cumulated attenuation along each ray, following the Beer-Lambert law. Everything the ray crosses is superimposed on the same image: a pore 2 mm below the surface and one at the centre of the part look identical.
CT breaks this ambiguity. The part rotates (or the source-detector system rotates around the part) and hundreds or thousands of projections are acquired from different angles over 360°. A reconstruction algorithm — typically filtered back projection, in its FDK variant for cone-beam geometries, or slower but noise-robust iterative algorithms — combines the projections into a 3D voxel volume, where each voxel represents the local attenuation coefficient of the material. The fundamentals, terminology and examination practices of the method are codified in public standards: the ASTM E1441 guide, the ASTM E1570 practice and the ISO 15708 series (radiation methods for computed tomography).
The practical outcome: flaws located and measured in 3D, densities distinguished, internal surfaces reconstructed and measured. It is the difference between "this casting has porosity" and "this casting has a 0.8 mm pore 1.2 mm below the machined surface, out of tolerance for the specification".
Architectures: cone-beam, fan-beam, microfocus and nanofocus
The two main acquisition geometries come with precise trade-offs:
- Cone-beam (cone-shaped beam + flat panel): acquires an entire volume per rotation, so it is fast and is today's dominant architecture in laboratory systems. The price is higher sensitivity to scatter and cone artifacts at large angles.
- Fan-beam (fan-shaped beam + line detector): acquires one slice at a time, is slower but far less sensitive to scatter thanks to collimation: it remains the preferred choice at high energies and on massive parts, and it is the reference architecture of the ASTM E1570 practice.
On the source side, classification by focal spot size determines the achievable level of detail: minifocus tubes for castings and large components, microfocus (spot of a few micrometres) for fine detail at high geometric magnification, nanofocus (below one micrometre) for microelectronics and materials. At very high energies, for massive components, linear accelerators replace tubes.
The parameters that drive quality: voxel, focal spot, kV
Three interlinked quantities govern the real resolution of a scan:
- Geometric magnification: M = source-to-detector distance ÷ source-to-object distance. Moving the part closer to the source magnifies the image on the detector.
- Voxel size: to a first approximation, the detector pixel pitch divided by M. Caution: the voxel is a geometric parameter, not the resolution. A 5 µm voxel with a 30 µm focal spot does not "see" 5 µm details.
- Focal spot: at high magnifications the geometric penumbra generated by the focal spot becomes the sharpness limit. This is why high resolution requires microfocus/nanofocus tubes, which however deliver low power: the resolution-power compromise is structural.
The choice of kV and power depends on the material and thickness to penetrate: indicatively 130-160 kV for polymers and electronics, 225 kV for light alloys and thin steels, 300-450 kV for thicker steels and castings, accelerators beyond. Attenuation grows with density and atomic number: coefficients are tabulated in public sources such as the NIST XCOM database, useful to estimate a priori the penetrability of a material at a given energy. Beam filtration (copper, tin) hardens the spectrum and reduces artifacts at the expense of contrast.
Typical artifacts and how to mitigate them
- Beam hardening: the polychromatic spectrum "hardens" as it crosses the material (low energies are absorbed first), producing the classic cupping effect and dark streaks between dense regions. Mitigations: physical filtration, software linearisation corrections, dual-energy in difficult cases.
- Scatter: radiation scattered by the part and the environment reduces contrast and grey-value accuracy, especially in cone-beam on dense materials. Mitigations: collimation, anti-scatter grids, software corrections, or a fan-beam geometry.
- Ring artifacts: concentric rings caused by non-uniform detector pixels. Mitigations: flat-field calibration, detector-shift motion, reconstruction filters.
- Cone artifacts: away from the central plane of the cone-beam geometry, FDK reconstruction is approximate. Mitigations: limiting the cone angle, helical trajectories, iterative algorithms.
- Metal and truncation artifacts: very dense inserts or parts wider than the field of view generate streaks and wrong values; they are managed with part orientation, higher energies and dedicated scanning techniques.
The applications that justify the investment
- Porosity and internal defects: quantitative analysis of pores, shrinkage cavities and inclusions in castings (aluminium high-pressure die casting above all) and additive-manufactured components: size, position, distance from the machined surface, porosity percentage per zone.
- Metrology and CAD-to-part comparison: from the surface extracted from the volume you run colour-map comparison against the CAD, GD&T dimensions and tolerances, wall thicknesses, measurements on inaccessible internal geometries. For metrological use of CT — uncertainty, influence quantities, performance verification — the public reference is the VDI/VDE 2630 series of guidelines.
- Assembly analysis: non-destructive verification of assembled products: O-ring position, plastic welds, clips, connectors, dosing, without disassembling or sectioning.
- Electronics: voids in BGA solder joints, wire bonding, delaminations, potential shorts: the home ground of microfocus and nanofocus systems.
CT, 2D radiography or ultrasonic testing? Decision table
| Criterion | Industrial CT | 2D radiography (RT/CR/DR) | Ultrasonic testing (UT/PAUT) |
|---|---|---|---|
| 3D flaw location | Yes, with volumetric measurement | No (superposition) | Partial (depth yes, limited imaging) |
| Quantitative porosity | Excellent (per zone and per pore) | Qualitative | Limited |
| Internal-geometry metrology | Yes (with stated uncertainty) | No | Thickness only |
| Oriented planar cracks | Good if orientation is favourable | Depends on orientation vs beam | Excellent (PAUT/TOFD) |
| Thick steel sections | Limited (high energies needed) | Good with adequate sources | Excellent |
| Time and cost per part | High | Medium-low | Low |
| Typical use | Process development, qualification, failure analysis, critical series | Weld and casting inspection in production | Welds, corrosion, in-service thickness |
For a deeper comparison between radiographic and ultrasonic methods, see also industrial radiography vs ultrasonic testing and the CR vs DR guide.
How to choose a CT system: criteria and questions to ask
- Maximum energy: it must cover the worst material/thickness in the part mix, not the average part. A 225 kV system at its limit on today's part will be insufficient on tomorrow's.
- Measuring volume and manipulator: maximum part size and weight, axis travel, motion accuracy (decisive for metrology).
- Required resolution vs part size: define the minimum flaw to detect and verify it is achievable with the magnification possible on that part, not on the brochure sample.
- Throughput: cycle time for scan + reconstruction + analysis; for series inspection, automation, recipes and automatic defect evaluation matter.
- Software: reconstruction, porosity analysis, CAD-to-part comparison, GD&T measurement, reporting: often worth as much as the hardware.
- Metrology: if the CT must issue measurements, ask for performance verification per VDI/VDE 2630 and the uncertainty determination strategy.
- Radiation safety, installation and service: shielded cabinet, licensing requirements, operator training, response times and spare-part availability in your country.
Common mistakes
- buying brochure resolution: the voxel quoted on the smallest sample is not the resolution on your real part;
- undersizing the energy with respect to the future material mix;
- ignoring analysis time: the scan is often the short part of the cycle;
- using CT for dimensional measurement without determining uncertainty;
- not testing your own sample parts before purchase, with declared cycle times;
- neglecting software, training and service when comparing offers.
PITECH supports the selection of industrial computed tomography and industrial radiography systems with a neutral approach: it starts from parts, materials, expected defects and required throughput, defines the technical specification and compares architectures on the real sample, not on the catalogue.
Frequently asked questions about industrial CT scanning
What is the difference between industrial radiography and industrial CT?
2D radiography produces a single projection in which all the material along the beam path is superimposed: a flaw is visible but cannot be located in depth. CT acquires hundreds or thousands of projections while the part rotates and reconstructs them into a 3D voxel volume: every flaw is located and measured in three dimensions, porosity is quantifiable and internal geometry is measurable. In exchange, CT takes longer per part and requires a higher investment.
What resolution can an industrial CT system achieve?
It depends on voxel size, focal spot and geometric magnification. Microfocus systems reach voxels of a few micrometres on small parts, nanofocus systems go below one micrometre. On large parts resolution degrades because magnification is limited by the field of view: as a practical order of magnitude, usable resolution scales with part size. Real resolution never equals the voxel size alone: focal spot, mechanical stability and the detector also count.
Which materials and thicknesses can industrial CT penetrate?
It depends on tube energy: indicatively 130-160 kV for polymers, electronics and thin light alloys; 225 kV for structural aluminium and thin steel; 300-450 kV for thicker steels and castings; beyond that, linear accelerators. Attenuation depends on density and atomic number, with coefficients tabulated in public sources such as the NIST XCOM database. On dense materials image quality degrades due to beam hardening and scatter well before the penetration limit.
Can industrial CT be used for dimensional metrology?
Yes: CT is the only method that non-destructively measures internal geometries inaccessible to touch probes or optical scanners. From the reconstructed volume a surface is extracted for CAD-to-part comparison, GD&T dimensions and wall-thickness analysis. Uncertainty must however be determined and stated: the VDI/VDE 2630 series defines fundamentals, influence quantities and performance verification of CT systems used as measuring machines.
When is it worth buying a CT system instead of using an external service?
A scanning service makes sense for occasional needs, feasibility studies and one-off qualifications. Buying becomes rational with recurring volumes (series inspection, foundry or additive process development, frequent failure analysis), when service turnaround becomes a bottleneck, or when parts cannot leave the plant for confidentiality. The evaluation starts from parts, materials, expected defects and required throughput.