Tomra GAINnext Uses Deep Learning Technology to Purify PET Streams for Recycling

TOMRA recycling is highlighting the latest advances in its artificial intelligence add-on, now rebranded as “GAINnext.” GAINnext uses deep learning, a machine learning technology that uses an artificial neural network to execute more effective sorting by recognizing objects by material and shape.

According to the company, GAINnext enables the sorting of objects that could not be separated by traditional means. Combined with TOMRA’s Autosort system, GAINnext classifies material based on sensor data and provides object recognition using an RGB camera.

A recent application of GAINnext aids in the creation of a recyclable PET fraction. The technology enables the separation of contaminants, such as multilayer packaging, to create a clean PET bottle stream. The PET cleaner application targets the separation of white opaque PET from clear and light blue PET, a difficult challenge with previous technology.

TOMRA has also announced applications for sorting food grade from non-food grade plastics, often difficult to distinguish, for PET, PE and HDPE materials. Efficient sortation of this material could improve material recovery.

TOMRA’s deep learning technology was first introduced in 2019 and has been implemented in more than 100 Autosort units at material recovery facilities worldwide. The initial application removed PE-silicon cartridges from PE waste streams.