For other resins with different cap material is expected that
the proposed system will distinguish the target resin properly.
These resins are not expected to cause any trouble in sorting
system, because the samples with considerable percent of cap
coverage are not often observed in real waste.
4. Surface contamination
Contaminated samples do not affect the quality of NIR
reflectance spectra, because in the NIR wavelength region the
length of light pass is much higher (about 10 mm at the second
CH stretch overtone about 1200 nm) than the layer thickness
of the dirt (< 0.5 mm) [17].
The selected samples for spectroscopy were not washed and
not cleaned. As such, some of them had a dirty surface.
However, after taking spectra, it was concluded that the
spectra of dirty samples are similar to spectra of clean samples
for a given resin (Fig. 10).
Fig. 10 The NIR reflectance spectra of clean and dirty blue PP
D. Cost Considerations
With all sorting categories available for plastic type and
color separation, the infrastructure and operational cost would
arise as a serious challenge for the success of an automatic
sorting system [8]. The proposed identification and
classification system represents a significant portion of the
costs associated with the recycling process. In the current
work, identification hardware and classification system
required costly component such as optical filter, DAQ card,
GRIN lens, face plate, optical fiber cable, and conveyor belt.
However, this method provided a higher value and accurate
sorted plastic output. Furthermore, sorting the plastic bottles
and containers with this method would result in the
considerably value added that legitimizes the costs of
identification hardware and sorting system in long term.
Furthermore, in other methods such as manual identification
and sorting, increasing labor costs are making manual sorting
economically unviable. In addition, the resulting product
which is costly produced, generally only finds limited