Whether an X-ray can detect a plastic foreign object depends on its chemical composition, not on the "plastic" itself. PVC and PTFE usually stand out clearly against food products, whereas PE and PP are barely more visible than fat. While dual-energy X-ray and AI improve detection, they cannot create contrast where none physically exists. Optical sorting and detectable plastics help to close this gap.
Plastic remains invisible if it attenuates X-rays to a similar degree as the food surrounding it. An X-ray scanner does not see materials, but rather differences in attenuation. This depends on the density and the atomic number , which is determined by the chemical elements a substance is made of. Absorption increases sharply with the atomic number—which is why metal and glass are so clearly visible.
Food consists of water, fat, and protein—that is, hydrogen, carbon, nitrogen, and oxygen—just like many plastics:
The rule of thumb "the denser, the more visible" is therefore too simplistic: according to NIST, PET (1.38 g/cm³) and PVC (1.406 g/cm³) have almost the same density. However, at 40 keV, PVC attenuates about 3.2 times as much as PET—solely because of the chlorine.
X-rays detect PVC well, PTFE usually, PET barely, and PE as well as PP practically not at all. We calculated the table using NIST reference data: it shows how much a 2 mm thick piece of plastic changes the X-ray signal compared to muscle tissue.
* PP has the same carbon-to-hydrogen ratio (CH₂) as PE and therefore the same mass attenuation coefficient; value derived from PE coefficient × PP density of 0.91 g/cm³ (APSOplast data sheet).
Key takeaway: PVC attenuates X-rays to a similar degree as glass, whereas PE and PP attenuate them less than the food product itself, resulting in a signal on the scale of a fat inclusion. Anyone wanting to detect plastic via X-ray must therefore first know which plastic is used in the production line.
Our calculations: μ/ρ and densities from the NIST Standard Reference Database 126; μ = μ/ρ × density; signal difference = 1 − e−|Δμ|·d for a 0.2 cm thick piece displacing meat. 40 keV simplifies the mean energy of a tube at around 80 kV; at 30 keV, the order remains the same. Tube spectrum, noise, detector resolution, packaging, and product thickness are not taken into account. The values show the physical upper limit of contrast, not the detection rate of a system—that can only be determined by a test under real-world conditions.
Dual-energy is particularly helpful for plastics that contain a heavier element such as chlorine or fluorine. The process compares absorption in two energy ranges; materials with a higher atomic number exhibit a different "energy signature" than the food product.
According to NIST values, between 30 and 50 keV, the mass attenuation coefficient of PVC drops by a factor of 3.3, that of muscle tissue by a factor of 1.7, and that of PE by only a factor of 1.3. This is why dual-energy X-ray from the RaymanTech Dual-Energy series distinguishes chlorine- and fluorine-based plastics, rubber with fillers, as well as bone and stone in non-uniform products better than single-energy X-ray.
This limit is confirmed by a study from the Fraunhofer Institute for Integrated Circuits IIS (Bauer, Wagner, Leisner, 2021): For frozen berries, vegetables, and ready meals (test specimens 0.3 to 10 mm), dual-energy outperformed conventional transmission in inhomogeneous products but did not detect plastics reliably.
Our verdict: For lines with PVC, PTFE, or filled rubber parts, dual-energy is a significant improvement. If the critical parts consist of PE or PP, dual-energy alone will not solve the problem.
No, AI recognizes patterns that a pure threshold analysis misses, but it cannot invent contrast. Trained models distinguish shape, edges, and texture, and can, for example, separate a sharp-edged PTFE splinter from a fat or cartilage structure and mask out packaging edges, clips, or weld seams. This reduces false rejects.
However, for PE or PP in meat, the signal difference according to our calculation is around 1.4 to 1.5 percent—in the same order of magnitude as the difference between fat and muscle (1.1 percent). If this difference is lost in image noise or the natural variation of the product, even the best AI cannot reliably identify it.
X-ray fails to find plastic when the material, shape, or product pushes the contrast below the detection limit—typically in these cases:
Where X-ray fails due to chemical properties, optical methods focus on the surface. Cameras detect color and shape, while NIR and hyperspectral systems additionally identify the material signature. For its DynaCQ vision system, the DTI reports the ability to detect plastic parts as small as 2 × 2 mm on a 70 cm wide belt at speeds of up to 70 m/min. The limitation: optics only see what is on the surface. This is not a disadvantage for single-layer bulk goods, but it is for packaged or thick products.
Often overlooked: Black plastics are invisible to NIR. According to a study in Polymers (Becker, Sachsenheimer, Klemenz, 2017), even small amounts of carbon black absorb all light in the NIR range. When using NIR sorting, you should avoid black plastic parts on the line.
For products running in a single layer, optical sorting from RaymanTech is the most sensible complement to X-ray; RaymanTech also offers combination systems featuring both X-ray and optics in a single machine. Further from practical application are millimeter waves: in 2020, the Fraunhofer FHR introduced a prototype called SAMMI, which is designed to detect plastic, glass, and wood, but cannot penetrate metal. It is not yet a standard for mass production.
On X-ray lines, X-ray detectable plastics are particularly helpful because they provide the contrast that normal plastics lack. Many companies are replacing tools, gloves, scrapers, cable ties, and belt components with detectable versions—two types are often confused:
The US Food and Drug Administration (FDA) classifies hard or sharp foreign objects between 7 and 25 mm as a health hazard (Compliance Policy Guide Sec. 555.425), and even those under 7 mm for products intended for infants, the elderly, or patients. For optical sorting, a high-contrast color such as blue is also recommended.
In the EU, plastic is the third most frequently reported foreign object. An analysis of the EU rapid alert system RASFF (Prata, Microplastics, 2024) counts a total of 102 reports of plastic foreign objects in food and feed between 2020 and 2023, ranging from 19 to 34 per year; only glass and metal were reported more frequently. Since many reports are misclassified, the actual number is likely higher.
In Germany, according to the Federal Office of Consumer Protection and Food Safety (BVL), 480 out of 2,295 warnings on lebensmittelwarnung.de between 2012 and May 2024 involved foreign objects, approximately 21 percent; in 2023, there were 46 out of 310. The BVL does not provide a breakdown by material.
The best approach is in three steps:
Which combination of X-ray, optical, and preventive measures makes sense for your line is shown on our pages on Foreign object detection in product inspection and to X-ray scanners for food. RaymanTech offers trial runs with your own products – request a trial run with your products and plastic parts.
Sie möchten wissen, wie sich das in Ihrer Linie umsetzen lässt? Auf der passenden Produktseite finden Sie Technik, Bauformen und Einsatzbereiche im Detail. Für eine Einschätzung Ihrer Anwendung sprechen Sie direkt mit unseren Inspektionsexperten.
Foreign Object Detection SolutionsOnly certain plastics. PVC and PTFE stand out clearly due to their chlorine and fluorine content, respectively. In many products, PE and PP are indistinguishable from grease, while PET is barely visible despite its high density.
Naturally PVC and PTFE. Other plastics achieve this through high-atomic-number additives, typically 5 to 30 percent by weight, according to Eriez. Whether a part can be detected on your line can only be determined by testing it with your product.
Standard plastic cannot be detected, as it is neither conductive nor magnetic. Metal-detectable plastics can only be detected if the pieces are significantly larger: in Vikan's testing, a piece approximately 11 mm in size was equivalent to a 1.5 mm iron ball.
Generally unreliable. PE consists only of carbon and hydrogen and its atomic number is barely distinguishable from that of fats and proteins; furthermore, thin films provide almost no signal. Optical systems or preventative measures are better suited for this purpose.
The FDA classifies hard or sharp foreign objects measuring 7 to 25 mm as a health hazard, with even smaller objects posing a risk to vulnerable groups. In the EU, there is no fixed size threshold; assessments are conducted on a case-by-case basis via HACCP hazard analysis.