The Industry 4.0 paradigm requires new technologies and methods not only to improve the profitability and the quality of the industrial production and products, but also new strategies to reduce the social and environmental impact of the production process. Many line manufacturing chains unbox and assembly components to create products, but create a large amount of waste that sometimes can't be recycled because of the exposure to contaminants. When it comes to the automotive industry, mineral oils may contaminate plastic packaging and cardboard boxes during manufacturing, making hard to recycle them. In this paper we propose a proof of concept of a packaging sorting system based on NIR spectroscopy, to automate sorting and get high quality outputs for the recycling of cardboard package boxes. Spectral datasets have been pre-processed and dimensionally reduced using PCA A SVM algorithm has been trained to distinguish between oil contaminated and non contaminated materials. Two NIR spectrometers with sensing range 640-1050 nm and 950-1650 nm have been used and evaluated, to select the proper sensor configuration. Eventually, the system classification accuracy was respectively up to the 98,68% and 98,64% using the 950-1650 nm and the 640-1050 nm spectrometers, demonstrating the opportunity to detect mineral oil contamination on boxes.
|Titolo:||In-line industrial contaminants discrimination for the packaging sorting based on near-infrared reflectance spectroscopy: A proof of concept|
CAVALLO, Filippo (Corresponding)
|Data di pubblicazione:||2017|
|Appare nelle tipologie:||4.1 Contributo Atti Congressi/Articoli in extenso|