Research On Plastics Classification Algorithms In Laser-Induced Breakdown Spectroscopy

Plastics have been applied in industries and daily life. To protect the environment and save resources, it is the best way to reuse and recycle post-consumer plastics. Plastics classification is a crucial step in the process of recycling. The existing methods, such as X-ray Fluorescence Spectroscopy, Near Infrared Spectroscopy and Raman Spectroscopy, have been well developed during the past few decades. However, due to their deficiencies, including complex samples preparation and time-consuming, it is difficult to meet the demands of plastics classification online in industry for these technologies. As a new classification technology, laser-induced breakdown spectroscopy(LIBS) with its advantages of high-speed, no or less sample pretreatment and in-site analysis is applied in many areas, such as polymers, metals, soil and ceramics etc.Focusing on fast classification of plastics by laser-induced breakdown spectroscopy, the effects of several key process parameters(laser pulse energy, ICCD gate and width) on spectroscopy emitted from polymers plasma have been studied and optimized based on the self-built LIBS system. 20 kinds of plastics with different colors from different manufacturers were studied on fast classification in open air. 6 characteristic spectral lines corresponding to carbon, hydrogen, oxygen and nitrogen were chosen to avoid the interference of metal spectra. The results showed that the average identification accuracy with support vector machine(SVM) was 96.6%, and the average identification accuracy was 99.6% when the plastics with different colors but the same type were classified into the same category.In order to improve the classification efficiency, unsupervised algorithms, including Principal Component Analysis(PCA), K-means and Self-organizing Maps(SOM) were studied, and U-matrix algorithm was used to visualize the result of 20 plastics classification by SOM. By comparing the classification results of the three algorithms above, the best result can be obtained by the SOM based on U-matrix algorithm. The above results indicated that LIBS combined with unsupervised algorithms could be used to classify various plastics with high efficiency and accuracy, which provides a new way for plastics fast classification and recycling.