InsightsSpectral recognition
How does NIR sorting work?
Direct answer
NIR sorting identifies what a piece is made of by measuring how it reflects near-infrared light. Every polymer returns a distinct spectral fingerprint across multiple wavelengths, so plastics that look identical to a camera are still separated by material. Each particle is classified and ejected within milliseconds, at production speed.
The three steps, in order
Near-infrared analysis identifies polymer type by measuring spectral response across multiple wavelengths. In a running machine that resolves into three steps, and each one has to hold for the result to be repeatable.
- Polymer signature recognitionEach polymer has a distinct spectral fingerprint, captured across the near-infrared range. This is the property the whole method rests on: the fingerprint belongs to the material, not to its color or its shape.
- Multi-wavelength analysisBroad spectral scanning separates plastics that look identical to the human eye or a color camera. A single wavelength would leave ambiguities that a wide scan resolves.
- Real-time classificationWithin milliseconds each particle is classified and routed, so sorting keeps pace with production speed. Recognition that cannot keep up with the flow is not sorting.
What NIR reads
- Polymer type by spectral signature — PET, HDPE, PP, PE, PVC, PS, ABS, PC.
- Signal shifts caused by additives, coatings and fillers.
- Composition conflicts such as moisture, ageing or contamination.
The last two lines are conditions, not target classes. NIR primarily reads polymer material response; additives, coatings, moisture, ageing and contamination shift or weaken that response rather than being separated out as their own fraction. They surface as a signal that does not match the validated library cleanly, which is why real samples and calibration are required — and why feed condition matters as much as sensor quality.
NIR and color sorting answer different questions
These are not competing technologies and one does not replace the other. They read different properties, so they fail in different places — which is exactly why a line often stages both.
| Aspect | Color sorting | NIR material sorting |
|---|---|---|
| Question it answers | What does this piece look like? | What is this piece made of? |
| What it reads | Color, transparency, shape and visible defects | Spectral response across the near-infrared range |
| Typically catches | Off-color flakes, black pieces, and foreign objects such as wood, metal, stone, rubber and paper | PVC hiding in clear PET; PP mixed into a PE stream |
| Cannot catch | A color-correct piece that is still the wrong polymer | Black and carbon-black pieces, which absorb the signal |
One consequence is easy to get backwards: NIR is for polymer identification, not for foreign objects. When the question is wood, metal, stone or paper in a plastic stream, the answer is optical color sorting — those things look clearly different from the plastic around them, and that visible difference is what gets detected.
What NIR needs to work well
- Light-colored and translucent plastics with clean, dry surfaces — this is where performance is best.
- Validated spectral libraries, so a measured fingerprint has something correct to be matched against.
- Stable material flow and consistent presentation, so every piece is actually inspected.
- Periodic calibration, because accuracy drifts if the reference does.
Feed condition does more of the work here than a specification sheet suggests. The same material arriving clean, dry and evenly presented behaves differently from the same material arriving wet, dusty or surging — which is why a configuration is built from a representative sample rather than from a material name.
Where this stops applying
- NIR cannot reliably identify black or carbon-black plastics, which absorb the signal. They can be removed as visible rejects by color sorting, but that does not identify their polymer type.
- NIR is polymer identification. It is not the route to removing wood, metal, stone or other non-plastic foreign objects — that is optical color sorting.
- Accuracy depends on validated spectral libraries, stable flow and periodic calibration, not on the sensor alone.
- Mayson may describe a reference reject-removal range in early discussion — under suitable material conditions and proper machine setup, at least about 90–98 out of every 100 target reject pieces. That is a reference range, not a guaranteed final product purity, and it does not replace sample testing.
- A single sorting machine does not guarantee food-grade or medical-grade output. That depends on the complete washing, decontamination, extrusion, testing and certification process.
Common follow-ups
What does NIR stand for?
Near-infrared. NIR sorting identifies polymer type by measuring how a piece responds to light in the near-infrared range, across multiple wavelengths, rather than by how it looks.
Can NIR tell PET from PVC?
Yes, and it is the priority job in PET recycling. Clear PVC can look almost identical to PET and pass color sorting, so NIR reads the polymer signature to separate accepted PET from PVC-rich and other non-PET reject classes.
Why can NIR not read black plastic?
Black and carbon-black plastics absorb near-infrared light, so no usable spectral fingerprint returns. The standard sequence is to remove black items first through optical color sorting, then run the material separation on what remains.
Does NIR work on whole bottles or only on flakes?
Both, on different platforms. MAS-P and MAS-P Pro apply NIR to prepared flakes; MAS-B Pro adds NIR polymer identification to whole-bottle sorting, so bottles are separated by material before shredding.
Is NIR sorting the same as a color sorter?
No. A color sorter classifies color, transparency and visible appearance and does not identify polymer type. NIR identifies what a piece is made of and is blind to black material. They read different properties, which is why lines often use both.