At the Biffa Polymers facility in Redcar, advanced optical and infra-red scanning technologies now analyze the composition of plastic waste moving along high-speed conveyor belts at rates that manual sorters simply cannot match.
Computer Vision Solves High-Density Polyethylene Crushing Challenges
When a standard translucent milk bottle is intact, legacy optical sensors identify its profile easily. Once crushed or flattened under the weight of municipal waste streams, however, its physical contours alter entirely.
According to Biffa Polymers managing director James McLeary, AI systems resolve this limitation by learning to recognize crushed geometries. “A milk bottle might be easy to recognise when it is intact, but once crushed its shape can change completely,” McLeary explained, noting that the computer vision models are trained to verify that irregular shapes still map to targeted polymer signatures.
Inside the Redcar Closed-Loop Mechanical Pipeline
Operating continuously for nearly two decades, the Redcar facility has processed the equivalent of more than 10 billion plastic milk bottles. The facility handles the final stages of a complex mechanical pipeline once municipal waste has undergone initial sorting by type and color.
- Shredding & Washing: Incoming plastic streams are shredded into flakes, washed thoroughly, and chemically separated.
- Extrusion: Cleaned flakes are treated and melted down into fine plastic pellets.
- Distribution: These tiny HDPE pellets ship to external manufacturing plants to produce food-grade packaging.
This automated triage dramatically compresses supply chain loops. As McLeary noted regarding the speed of modern processing, a “bottle that you threw away two weeks ago may be back in somebody else’s fridge, because that’s how quickly the material gets recycled in the UK.”
Balancing Workforce Evolution With Advanced Facility Safety
Far from operating as an entirely lights-out facility, the plant maintains a workforce of approximately 200 people. Rather than replacing human labor entirely, machine learning models absorb tedious, repetitive sorting duties. This allows personnel to focus on quality assurance and verifying that final output meets strict food-grade standards.
Beyond material classification, the facility utilizes computer vision algorithms for workplace safety. Cameras monitor operational zones across the site, flagging potential risks in real-time—such as instances when a human operator moves into an active vehicle path.
Evaluating Future Robotic Integration Limits
Looking ahead, the company is investigating additional robotic sorting systems capable of physically extracting targeted bottles from the stream using integrated machine learning classifiers. Nevertheless, management emphasizes that the heavy physical lifting of line sorting still relies on human workers prepared to handle manual responsibilities.
