Industrial polymers classification using laser-induced breakdown spectroscopy combined with self-organizing maps and K-means algorithm

To extend the industrial polymer species classification and improve its efficiency. Laser-induced breakdown spectroscopy (LIBS) combined with unsupervised learning algorithms of self-organizing maps (SOM) and K-means was employed to differentiate industrial polymers in the open air. Only the intensities of non-metallic lines, including two molecular band lines (C-N(0,0) 388.3 nm and C2(0,0) 516.5 nm) and four atomic emission lines (C I 247.9 nm, H I 656.3 nm, N I 746.9 nm and O I 777.3 nm) were used. Firstly, the SOM neural network with adjusting spectral weightings (ASW) was applied to separate 20 kinds of polymers preliminarily. The results were obtained in the output space which indicated that 18 kinds of polymers have been separated except for polycarbonate (PC) and polystyrene (PS). Afterwards, the K-means clustering algorithm was utilized to separate PC and PS. The accuracy of the industrial polymers classification for 20 kinds of polymers was 99.2%. It demonstrated that the feasibility of clustering of industrial polymers using LIBS.