Identification and Sorting of Plastics Film Waste by NIR-Hyperspectral-Imaging

INTRODUCTION
Plastics films are widespread products used by the packaging sector, in form of bags or protective foils, which need to be properly managed when they reach their end of life phase in order to protect the environ-ment and save valuable resources. As any other waste they must be classified and sorted in groups with the same chemical composition be-fore recycling, but due to their low density and geometry they are more difficult to sort. In addition to classical methods of material sorting, these industrial operations are increasingly supported by optical sensor technology, especially NIR-Hyperspectral-Imaging (NIR-HSI). In the present contribution the possibilities and limitations of an industrial NIR-HSI-based film sorting machine are reviewed.
EXPERIMENTAL
A standard sorting machine of the type UniSort PR from RTT Steinert has been expanded with guide shafts for the suppression of turbulent air flows on the conveyor belt. A hyperspectral camera (called RTT-HSI 2.2) has been used as optical sensor. Prior to real time experiments a spectra data base has been generated off-line, that includes spectra of all hereby rele¬vant kinds of chemical plastics film compositions. Of par-ticular interest were the so-called bio-plastics because of their above-average growth on the mar¬ket. Sets of different “classical” (PP, LDPE) and bio-plastic films (starch, PLA, oxo-PE) were collected and spectro-graphically measured for the development of robust classification pro-cedures. The results were adopted into the on-line Hyperspectral-Imager of industrial scale (RTT-HSI 2.2). Then a larger number of sort-ing experiments were carried out. The details of the sorting trails are not subject of the present contribution since they are mainly of technical nature. Only the spectrometric aspects are mentioned in the result discussions. The main problems when sorting films are, in addition to the material handling and transportation, the material overlapping and the between clustering. For the decomposition of information RTTs proprietary software was used.
RESULTS AND DISCUSSION
The Hyperspectral-Imager RTT-HSI 2.2 has been tested in industrial scale with regard to the separation of different types of plastics films. As basic machine a modified sorter of type UNISORT PR has been used. The input stream was a waste mix of fossil origin and bio-based plastics, LDPE and PLA-films respectively. LDPE- and PLA- films may not jointly processed in an extruder for their recycling due to their different physical properties and incompatibility. This would lead to a clumping of the melt. The sorting tests have shown that the considered plastics films can be exact separated from the input waste stream by Hyperspectral-Imaging in the 1.4 … 2.1 µm wavelength range. In conjunction with digital image processing and discriminant analysis, HSI allows automatic sorting of films in real time under industrial conditions.