AI Methods for Improving Polymer Sorting

The exponential rise in the production and use of polymers, particularly in single use applications, has led to a dramatic increase in their environmental prevalence and problems with waste management. Our waste system is built around mixed polymer waste, which makes separation a key challenge for increasing the value and recyclability of this waste. Here, we describe the development of different classification algorithms that enable new capabilities in mixed plastic waste sorting. First, a spectral classifier is developed that enables the sorting of different polyesters. The goal of this classifier is to enable emergent new bioplastics and biodegradable polyesters to be included within mixed plastics recycling streams. This classifier demonstrates greater than 95% efficacy at separating similar polymers such as poly(ethylene terephthalate) and poly(lactic acid). Second, machine vision is explored as a method to separate different polymers. In this case, the high recognizability of consumer packaging provides a method for differentiation of different plastics, allowing labelling to be used as an effective proxy for plastic type. This provides a useful method for achieving scalable separation consistent with extended producer responsibility.