Raman spectra unmixing to identify waste polymers

This paper explores advanced methodology for special spectral decomposition into components (endmembers) corresponding to real chemical components in the mixture.. The multivariate linear technique based on principal component analysis (PCA), realized by singular value decomposition (SVD) and subsequent transformation (rotation) by independent component analysis, is used for dimensionality reduction and transforming PCA components to meaningful endmembers. Ensuring that the extracted endmembers correspond to real chemical components in the mixture, the statistical independence among extracted endmembers and their constraints to be non-negative with their sum should be fulfilled. This problem is solved using constrained quadratic programming by the Newton method. The proposed algorithm is validated by demixing simulated spectra containing four components. The RAMIX program is written in the Python language, which is used for the analysis of simulated and experimental RAMAN spectra of polymeric mixtures. The reconstructed concentrations are compared with the true original concentration, with very low differences. An example of premortem plastics mixture waste chip Raman spectra analysis shows the usefulness of this approach to analyzing real polymeric mixtures.