Microplastic detection and recognition system enabled by a triboelectric nanogenerator and machine learning techniques

Abstract

The toxicity of microplastic pollutants is closely associated with their material, size, and concentration. However, current detection methods are plagued by issues such as high cost, long processing times, and inadequate database coverage. Therefore, this study designed and developed a simple and rapid detection and identification device for microplastics in water, based on the combination of liquid–solid contact electrification and machine learning algorithms. The study shows that the average open-circuit voltage difference among different types of microplastics ranges from 1.6 V to 11.7 V, and the average peak open-circuit voltage difference ranges from 1.7 V to 22.0 V; the Random Forest (RF) model achieved an average recognition accuracy of 95.24%. This work provides a new method for real-time online monitoring of microplastics, which holds significant implications for the environmental monitoring, food safety, medical and health fields.