Compostable plastic can offer a sustainable alternative for conventional plastics, especially in single use packaging, bags and bio-waste bags - but only if it is disposed of in the correct manner, so that it lands in the right waste stream.
This, however, does not always happen, which means that not only are the compostability benefits lost, these materials also end up contaminating the recycling stream of other plastics - the reason many recyclers remain utterly opposed to their use.
Now, researchers at University College London have developed technology to improve the accuracy of the sorting systems used today in order to eliminate these and other contaminants from the plastics waste stream.
In a paper published in Frontiers in Sustainability they describe how they used machine learning to automatically sort different types of compostable and biodegradable plastics and differentiate them from conventional plastics.
Following their evaluation of a number of different sorting technologies, the researchers identified hyperspectral imaging (HSI), which combines imaging technology and spectroscopy into one approach, as the most effective, non-destructive technique to use.
By applying shortwave infrared (SWIR) in the range 950–1,730 nm they could identify not just different types of conventional plastics (PP, PET, and LDPE) and compostable plastic (PLA, PBAT) packaging but also compostable materials (palm leaf and sugarcane-based materials) with various sizes from 50 x 50 mm to 5 x 5 mm.
However, the amount of spectral information collected by HSI from the sample surfaces that must be processed in order to make sorting decisions in real time is huge, which led to the decision to try a novel approach using machine learning methods to differentiate and classify the different materials.
A training dataset and a testing dataset were compiled. The training dataset was the dataset used to build the classification model. It was an input into the machine learning algorithms to allow the model to associate spectral imaging data with known material classifications. The testing dataset was the dataset that contains unseen data to test the model accuracy in determining material classifications. It was used to evaluate the performance of the model.
The researchers worked with different types of plastics including PP and PET, as well as LDPE. Compostable plastic samples included PLA and PBAT, Results showed that ‘the accuracy is very high and allows the technique to be feasibly used in industrial recycling and composting facilities in the future,’ said Mark Miodownik, corresponding author of the study.
The model achieved perfect accuracy for all materials when the samples measured more than 10mm by 10mm. For sugarcane-derived or palm-leaf-based materials measuring 10mm by 10mm or less, however, the misclassification rate was 20% and 40%, respectively.
Looking at pieces measuring 5mm by 5mm, some materials were identified more reliably than others: For LDPE and PBAT pieces the misclassification rate was 20%; and both biomass-derived materials were misidentified at rates of 60% (sugarcane) and 80% (palm-leaf). The model was, however, able to identify PLA, PP and PET pieces without error, regardless of sample measurements.
“Our system is capable of accurately sorting compostable plastics at the typical product scale (compostable spoons, forks, coffee lids) and differentiating them from identical looking conventional plastic items with high accuracy,” the researchers write.
However, for the system to be adopted by industrial composters, the classification speed needs to be increased to match the conveyor speeds in use, and real-time robotic removal of the plastics needs to be demonstrated, they added.