CT Scanning of Plastic and Composite Parts: What It Shows and How to Prepare the Analysis

At 60 keV, a typical energy for scanning plastic parts, half the photons get through about 3.4 cm of water, just over 9 mm of aluminium and under half a millimetre of copper. Those are NIST attenuation coefficients turned into thickness. Polymers made of carbon, hydrogen and oxygen behave roughly like water. That transparency is what makes CT so effective on a moulded pump housing, a connector or a carbon fibre laminate. The same ratio explains why a brass insert a few millimetres long, buried in that connector, can spoil the image of everything around it.

Quality engineers at a moulder or a composites shop ask more of CT than a foundry does: besides voids, fibre orientation, warpage against the tool, whether a weld has taken, where an insert ended up. CT answers some of these well and others only under certain conditions. One or two it hardly answers at all, and it pays to know which before shipping the part.

Voids Show Up Clearly in Polymers; Carbon Fibre Doesn't

CT contrast comes from the difference in attenuation between neighbouring regions. Between a polymer at around 1 to 1.4 g/cm³ and the air in a void the difference is sharp. With the low tube voltages plastics allow, a pore shows up even when it's only a few voxels wide (voxels are the small cubes that play the role of pixels in a 3D volume). The exceptions are polymers containing heavier elements: the chlorine in PVC, the fluorine in fluoropolymers, mineral fillers and brominated flame retardants. They raise attenuation and call for more energy, at the expense of contrast.

Achievable detail follows the rule that applies to every part. When the sample fills the detector, the voxel size is roughly the imaged diameter divided by the number of pixels. A connector 20 mm across on a 2,000-pixel detector scans at about 10 µm. A housing 200 mm across scans at about 100 µm. The part sets the limit more than the machine does.

Trouble starts when two materials of similar density have to be told apart. Carbon fibre and epoxy are both mostly carbon and attenuate similarly. Individual fibres, a few micrometres thick, only separate at even smaller voxel sizes, which means coupons a few millimetres wide. It's no accident that the review by Garcea, Wang and Withers in Composites Science and Technology (2018), devoted to CT of polymer composites, treats achieving contrast and the trade-off between sample size and resolution as central topics. Glass is denser and stands out from the resin far more easily. On a whole part, then, CT sees voids and glass fibre distribution well. For carbon fibre it mostly returns the average orientation of plies and tows.

In Injection Mouldings CT Finds Voids and Warpage, Not Weld Lines

The defect CT finds most reliably in a moulding is the shrinkage void in a thick section. It shares a cause with the sink mark: volumetric shrinkage that the packing phase didn't compensate. Where the skin is still thin it gives way and a sink forms. Where it has already set, the melt pulls away inside and leaves a void. CT gives its position, volume and shape, and the position tells you where to act: the gate, the packing pressure or the rib design.

The second application is comparison against CAD. The surface extracted from the volume is aligned to the nominal model, and a colour map shows where the moulding has too much or too little material, internal dimensions included. At first sampling of a new tool it's the fastest way to know what to correct. Polymers, though, change with their environment. Polyamides absorb moisture and grow, a plastic's thermal expansion coefficient is many times that of steel, and semi-crystalline grades keep shrinking after moulding. A comparable measurement therefore needs a conditioned part, for instance to the standard atmospheres in ISO 291:2008, and a stated temperature. ISO 1:2022 sets 20 °C as the reference temperature for geometrical specifications. The uncertainty of an individual dimension is determined as for metals, following guideline VDI/VDE 2630 Blatt 2.1:2015-06, as described in the article on CT dimensional metrology of internal geometry.

The weld line, where two flow fronts meet, is where expectations should come down. In an unfilled, well-fused polymer the density along the line is practically identical to the rest of the part, and CT won't see it. It becomes visible when it carries something else with it: a V-notch at the surface, micro-voids of trapped air or, in short-fibre compounds, an abrupt change in fibre orientation. Hannesschläger and co-authors (Case Studies in Nondestructive Testing and Evaluation, 2015) characterised the weld line of a short-fibre injection moulding in exactly this way, telling it apart from regions of different orientation even at 43 µm voxels. Joint strength, which is what actually matters, remains a matter for mechanical testing.

Short-fibre orientation is one of the most requested results. In a glass-filled PA66 it governs local stiffness and strength, and it's used to validate filling simulations. Bernasconi, Cosmi and Hine, in Composites Science and Technology (2012), compared the optical method on polished sections with micro-CT on a notched plate in short-fibre polyamide. Agreement was excellent only when the right section plane was chosen: good on a longitudinal section, weaker on a transverse one. The optical method depends on where you cut; CT measures orientation through the whole volume without cutting. That's how the layered structure typical of mouldings is read, with fibres aligned to the flow near the walls and more random or transverse in the centre.

In Composites, the Void Percentage Depends on How It's Counted

In laminates, matrix porosity is the defect that weighs most on resin-dominated properties. The review by Mehdikhani and co-authors in the Journal of Composite Materials (2019) covers five decades of research on how voids form, how they're characterised and how they affect interlaminar shear, tensile, compressive and flexural strength, toughness and fatigue. The common thread is that pore shape, size and location matter too, and a single percentage can't describe them.

The traditional methods are destructive. ASTM D2734-23 goes through burning off the resin and applies only where the effects of ignition on the material are known. ASTM D3171-22 removes the matrix by digestion, ignition or carbonization. ISO 14127:2024 combines densities with fibre mass after combustion or acid digestion, and adds a thickness method and a microscopy method. The first ones give an average per specimen; microscopy shows the pores, but on one section only. None tells you where the voids sit in the whole part, between plies or inside a tow, or whether they cluster near a radius. CT does, without destroying it. Little, Yuan and Jones (NDT&E International, 2012) compared the density method, microscopy and micro-CT on carbon fibre laminates. The density method depended on accurately known constituent properties and mass fractions, which gave large standard deviations. A more recent study by Elkolali and co-authors (Polymers, 2022) found that micro-CT detected voids more consistently, while acid digestion remained better for fibre and matrix percentages.

In CT, though, the value still depends on the threshold that separates void from resin and on the voxel size, below which the smallest pores vanish. That's the point to write into the order. No standard says how void content in composites should be counted in CT. Voxel size, thresholding method, analysed region and minimum counted pore volume are set by agreement between the parties, exactly as for metal castings in the case of two labs reporting two porosity values on the same part. The acceptance limit is written by the drawing or the customer's specification.

Geometry imposes a second limit. Conventional CT rotates the part in the beam, and a large flat panel scans poorly because the beam crosses it over very different lengths depending on the angle. Such parts are handled on cut coupons, on regions of interest or with limited-angle techniques such as laminography, while ultrasonic inspection remains the production tool. Closed delaminations, with both faces in contact, have almost no air gap and slip past CT. Industrial ultrasonic testing is very sensitive to planar flaws parallel to the surface.

A Brass Insert in a Connector: Why Metal Smears the Image

An X-ray tube emits photons across a range of energies, and a dense material strips out the low-energy ones first. The beam leaving the insert therefore has a higher mean energy than the beam leaving the plastic next to it. The reconstruction algorithm assumes a monoenergetic beam and misreads the difference. Brooks and Di Chiro described the effect in Physics in Medicine and Biology back in 1976. It's beam hardening, and it produces dark bands between metal parts and a false grey-level gradient inside the metal itself. Along the direction where the metal path is longest, so few photons arrive that noise takes over, and streaks cut through the surrounding plastic.

The countermeasures get more expensive as you go. The first is free: orient the part to shorten the longest metal path, for example with the insert axis parallel to the rotation axis. The second is a copper or tin filter in front of the tube, which pre-hardens the beam at the cost of some plastic contrast. Software corrections reduce the artefact without removing it, and on parts to be measured they leave doubt about the surfaces near the metal.

The choice depends on what needs to be known. If the question concerns the plastic around the insert, such as voids at the interface of an insert moulding, the metal image is sacrificed. If it concerns the position of the insert or a contact, energy goes up and a noisier plastic is accepted. When both questions matter equally, two scans each set for its own material often deliver more than a single compromise.

Plastic Welds and Bonded Joints: The Void Is Visible, Joint Strength Isn't

Hot plate, vibration, ultrasonic or laser welding: to CT the weld seam is just another region of plastic, and what it sees are density differences. It shows voids and porosity in the seam, the shape of the flash, stretches where the two parts never touched and a film of air remains, and misalignment of the half-shells. It doesn't see a joint where the surfaces touch but the polymer chains never interdiffused. The density is the same and the volume shows nothing, even if the joint fails at the first burst test. Bonded joints work the same way. Missing adhesive and bubbles in the bead are visible; weak adhesion isn't.

That's why CT of a plastic weld earns most alongside destructive testing. Scan the parts before the burst test, compare the failure point with the volume, and you learn which visible defect really matters. From there CT can become a credible sampling check.

The Question, the Part and the Conditions It's Scanned In

Everything starts with the question, because voxel size, energy and orientation are chosen around it; with several questions, one scan set for the finest detail is often enough. Then the parts. Send at least two from the same batch, one good and one rejected: the difference between the two volumes says which defect matters better than any threshold written in advance. Metrology needs conditioned parts, the moulding date and the CAD model, and no metal packaging in the beam. A good starting point is the CT scanning service, where feasibility is assessed before the quote, or look at the section on plastics, composites, electronics and additive manufacturing. When scans on the same part family become daily, at every tool sampling for instance, the conversation moves to custom industrial CT systems, with the microfocus source and detector chosen on the most critical part.

What to Send Us for an Evaluation

To tell you whether CT answers your question and how to set it up, we need a few precise details:

• photos or a drawing of the part, with overall size and weight;
• material: base polymer, type and percentage of filler or fibre, any metal inserts;
• the question: voids, CAD comparison, fibre orientation, welds, position of inserts or contacts;
• for metrology: CAD model (STEP) and drawing with tolerances, moulding date, conditioning;
• for composites: process (autoclave, infusion, RTM), layup if known, required void limit;
• number of parts, including one good and one defective sample;
• acceptance criterion or customer specification, if one exists.

You can send them by email to info@pitech-solution.com or on WhatsApp, with photos, drawings or CAD attached.

On polymers and composites, CT answers questions of geometry and voids well, and questions of adhesion and strength poorly. If your question sits in the first half, the next step is a scan of the real parts through our CT scanning service, with a quote in 24 hours. If it sits in the second, CT helps decide where to run the mechanical test, and industrial computed tomography should be paired with other methods from the start.