The study was addressed to investigate the possibility to apply hyperspectral imaging (HSI) techniques for on-line sorting or quality control of different types of plastic waste particles in industrial recycling plants, with particular reference to polyolefins from complex wastes. An HSI based architecture, working in the near infrared region (1000-1700 nm), has been set-up at laboratory scale to analyse virgin polyethylene (PE) and polypropylene (PP), different polyolefin waste samples and different contaminants (wood, rubber, foam, glass, aluminium, etc.). The spectra of virgin polyolefins have been used as reference standard samples. Two complex plastic waste streams have been investigated: automotive shredder residue and building and construction waste. The reflectance spectral signature of the different plastic waste streams have been acquired, analysed and compared. The differences in the spectral response of the two polymers have been clearly identified. Based on the analysis of HSI spectra, it was evident the possibility to define specific parameters useful for recognition of the two polyolefins, as for example a wavelength band ratio. In addition, principal component analysis has been also applied both to classify PP and PE samples and to characterize PP-PE stream from contaminants. It is important to outline that the best and most precise recognition logic should be based on more sophisticated and complex statistical analyses, such as Principal Component Analysis (PCA), Partial Least Square (PLS), Neural Network (NN), etc., that usually require long computation time. Considering that in most industrial applications the fast response of the detecting/sorting device is one of the main constraints, as for example when particles are moving on a conveyor belt and they are sorted on-line, the adoption of simplified logics, obviously working properly, is preferred. Such simplified logics, based on the selection of a limited number of wavelengths, not only decrease computation time, but also decrease the costs of the device. The achieved results can be utilised to define sorting and/or control logics to be applied in “on-line” architectures directly working at industrial plant level.
Waste polyolefin streams characterization by hyperspectral imaging / Serranti, Silvia; Bonifazi, Giuseppe. - ELETTRONICO. - (2010), pp. 10-10. (Intervento presentato al convegno IASIM-10 tenutosi a Dublin, Ireland nel 18-19 November 2010).
Waste polyolefin streams characterization by hyperspectral imaging
SERRANTI, Silvia;BONIFAZI, Giuseppe
2010
Abstract
The study was addressed to investigate the possibility to apply hyperspectral imaging (HSI) techniques for on-line sorting or quality control of different types of plastic waste particles in industrial recycling plants, with particular reference to polyolefins from complex wastes. An HSI based architecture, working in the near infrared region (1000-1700 nm), has been set-up at laboratory scale to analyse virgin polyethylene (PE) and polypropylene (PP), different polyolefin waste samples and different contaminants (wood, rubber, foam, glass, aluminium, etc.). The spectra of virgin polyolefins have been used as reference standard samples. Two complex plastic waste streams have been investigated: automotive shredder residue and building and construction waste. The reflectance spectral signature of the different plastic waste streams have been acquired, analysed and compared. The differences in the spectral response of the two polymers have been clearly identified. Based on the analysis of HSI spectra, it was evident the possibility to define specific parameters useful for recognition of the two polyolefins, as for example a wavelength band ratio. In addition, principal component analysis has been also applied both to classify PP and PE samples and to characterize PP-PE stream from contaminants. It is important to outline that the best and most precise recognition logic should be based on more sophisticated and complex statistical analyses, such as Principal Component Analysis (PCA), Partial Least Square (PLS), Neural Network (NN), etc., that usually require long computation time. Considering that in most industrial applications the fast response of the detecting/sorting device is one of the main constraints, as for example when particles are moving on a conveyor belt and they are sorted on-line, the adoption of simplified logics, obviously working properly, is preferred. Such simplified logics, based on the selection of a limited number of wavelengths, not only decrease computation time, but also decrease the costs of the device. The achieved results can be utilised to define sorting and/or control logics to be applied in “on-line” architectures directly working at industrial plant level.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.