Mechanical recycling of post-consumer polyethylene terephthalate converts bales of bottles and thermoformed trays into a heterogeneous regrind whose processability cannot be reduced to intrinsic viscosity or melting point alone. In bottle-to-bottle extrusion, sheet production, and fibre spinning, the flake stream is simultaneously controlled by colour-sort yield, residual moisture distribution, and poured bulk density. These three variables interact: optical sorters remove dark and opaque fractions that would otherwise shift bulk packing behaviour, while surface water on washed flake increases interparticle adhesion and can depress flow through drying hoppers and feed throats. Acceptance limits for PET flakes and regrind are therefore written not as single values but as paired measurement conditions with defined test methods, including ISO 15512:2019, ASTM D6869, ISO 60:1977, and ASTM D1895-17. Without a controlled combination of colour sorting, moisture removal, and bulk density management, lot-to-lot variation produces melt-pressure instability and hydrolytic degradation on production-scale equipment.

What Constrains Optical Sorting of PET Flake Beyond Simple Colour Detection?

Colour sorting of PET flake on production-scale equipment combines visible-spectrum RGB cameras with near-infrared reflectance sensors operating across 900–1700 nm. In chute-fed and belt-type sorters, throughput per metre of working width ranges from 2 t/h to 8 t/h for flake fractions of 8–12 mm. The control system separates clear flake from coloured flake when the CIELAB colour difference exceeds a threshold, commonly configured at ΔE*ab of 3.0 against a clear bottle-flake reference. Validation of the colour channel is performed according to ISO 11664-4:2008. The sorting rule is not purely colorimetric. Dark blue, dark green, and carbon-black pigmented flakes absorb strongly in the near-infrared region and are frequently misassigned if only NIR signal is used; visible-spectrum cameras are required to reject these fractions from a clear rPET stream. Production lines running mixed bales report reject rates of 6–15% when the input contains both clear and coloured bottle fractions; published data for a specific bale composition is limited.

High-speed pneumatic ejection introduces a mechanical constraint. Air ejector valves with response times of 0.5–1.5 ms must actuate at belt speeds above 3 m/s; at these conditions, valve latency translates into spatial displacement of 4–8 mm, making the effective reject boundary larger than the detected flake area. This effect is managed by splitting the flake curtain into lanes and by derating throughput when the feed contains high percentages of small, thin flakes. False rejects are observed when water droplets or label fragments are misread as coloured defects, particularly if the flake surface moisture has not been reduced below 0.3 wt% before optical inspection. X-ray fluorescence sensors are sometimes placed in parallel to reject chlorine-containing flakes by chlorine Kα emission, with clear-stream acceptance commonly set below 50 mg/kg total chlorine. The colour-sort stage therefore depends on the upstream washing and dewatering steps and on the downstream moisture control loop.

Residual moisture in PET regrind is a two-phase measurement problem. Mechanically dewatered flake leaving a centrifuge often carries 0.5–1.0 wt% surface water, which is visible in the bulk and causes bridging in dryers and feed hoppers. Bound water inside the amorphous phase must be removed by diffusion; for a 4 mm flake, the effective diffusion path is short enough that hopper residence times of 4–6 h at 160–170°C with a dryer dew point of -40°C are adequate to reach a total moisture content below 50 mg/kg. Karl Fischer titration by ISO 15512:2019 or ASTM D6869 is used for absolute moisture determination; loss-on-drying at 160°C can overstate water because volatile acetaldehyde and low-molecular-weight oligomers are evolved simultaneously. A calibrated NIR moisture probe placed after the dryer can track changes of 20 mg/kg against the primary Karl Fischer value, but probe accuracy is sensitive to flake colour and surface oxidation.

Amorphous flake from thermoformed scrap requires pre-crystallisation before high-temperature drying. If the crystallinity measured by differential scanning calorimetry according to ISO 11357-3 is below 25%, flake can agglomerate in a hopper dryer above 160°C. A fluidised-bed pre-crystalliser operating at 140–150°C for 20–40 min raises crystallinity above 35% and prevents bridging during subsequent drying.

When Melt-Phase Hydrolysis Couples Moisture To Intrinsic Viscosity

At melt temperatures of 270–290°C, water reacts with ester linkages in a pseudo-first-order hydrolysis reaction. Bottle-grade PET with an intrinsic viscosity of 0.80 dL/g measured by ISO 1628-5 can lose more than 0.05 dL/g during extrusion when the pre-dried moisture content exceeds 80 mg/kg. The loss is not uniform across the extruder; it accelerates in zones where the melt seal is incomplete and atmospheric moisture can contact unsealed regrind. Twin-screw compounding machines with L/D ratios of 40:1 are more vulnerable to local IV loss than short single-screw extruders because the longer residence time distribution broadens the hydrolysis window. Pre-drying is mandatory when the ambient relative humidity exceeds 60%, as dried flake re-exposed to 23°C and 50% RH regains moisture rapidly through surface adsorption. Published data for this specific configuration is limited.

Bulk Density And Feed-Throat Mass Transfer

Bulk density measurements separate flake lots by particle geometry and packing behaviour. The poured bulk density of post-consumer bottle flake is typically 0.30–0.45 g/cm³ as measured by ISO 60:1977; regrind from thick-walled injection-moulded scrap with a flake size of 2–4 mm can reach 0.62–0.75 g/cm³ under the same procedure. Mechanically compacted bulk density values are commonly 5–20% higher because vibration eliminates bridging and aligns irregular particles. However, neither ISO 60:1977 nor ASTM D1895-17 provides a harmonised tapped-density procedure for plastic flakes; such values are therefore indicative and not used for lot acceptance. The feed throat of a flood-fed single-screw extruder acts as a mass-flow meter only when the flake bed density remains constant; a reduction in bulk density from 0.42 g/cm³ to 0.35 g/cm³ has been observed to reduce mass throughput by 8–10% at constant screw speed in a 75 mm extruder. Published data for a specific configuration is limited.

Bulk density is not to be confused with true density. Amorphous PET has a true density of approximately 1.33 g/cm³ and crystalline PET approximately 1.45 g/cm³, as determined by ISO 1183-1. A low poured bulk density therefore reflects voids and unfavourable flake geometry rather than a change in polymer density.

Bulk density variability is not independent of colour-sort performance. Dark and opaque flakes removed from a clear rPET stream often derive from thick-walled injection scrap or barrier bottles, which increases the density of the reject stream and leaves the accepted clear stream with a lower bulk density and a narrower particle-size range. A clear flake lot with a poured bulk density below 0.35 g/cm³ can still be processed if gravimetric feeding and hopper agitation are used, but batch-to-batch variation greater than ±5% from the extruder calibration value triggers bridge formation and screw starvation. Sieve retention by ASTM D1921 is used to connect bulk density to flake geometry; a high fraction of particles above 10 mm lowers bulk density while increasing the probability of feed-throat bridging. Production-scale handling therefore couples optical sorting, drier residence time, and flake-size classification into one acceptance chain. A lot that passes colour sorting but fails the bulk density check can shift a 75 mm single-screw extruder into melt-pressure fluctuation of ±1.5 bar even when the moisture specification is met. Such intermittent feeding causes local overheating and acetaldehyde generation in barrier bottle preforms. The process window is maintained only when the three parameters are measured together and when the flake is processed within 4–6 h of drying.