TY - JOUR

T1 - Developing a prediction model in a lightweight packaging waste sorting plant using sensor-based sorting data combined with data of external near-infrared and LiDAR sensors

AU - Schlögl, Sabine

AU - Kamleitner, Josef

AU - Kröll, Nils

AU - Chen, Xiaozheng

AU - Vollprecht, Daniel

AU - Tischberger-Aldrian, Alexia

N1 - Publisher Copyright: © The Author(s) 2024.

PY - 2024/4/30

Y1 - 2024/4/30

N2 - Sensor-based material flow monitoring allows for continuously high output qualities, through quality management and process control. The implementation in the waste management sector, however, is inhibited by the heterogeneity of waste and throughput fluctuations. In this study, challenges and possibilities of using different types of sensors in a lightweight packaging waste sorting plant are investigated. Three external sensors have been mounted on different positions in an Austrian sorting plant: one Light Detection and Ranging (LiDAR) sensor for monitoring the volume flow and two near-infrared (NIR) sensors for measuring the pixel-based material composition and occupation density. Additionally, the data of an existing sensor-based sorter (SBS) were evaluated. To predict the newly introduced parameter material-specific occupation density (MSOD) of multi-coloured polyethylene terephthalate (PET) preconcentrate, different machine learning models were evaluated. The results indicate that using SBS data for both monitoring of throughput fluctuations caused by different bag opener settings as well as monitoring the material composition is feasible, if the pre-set teach-in is suitable. The ridge regression model based on SBS was comparable (RMSE = 3.59 px%, R² = 0.57) to the one based on NIR and LiDAR (RMSE = 3.1 px%, R² = 0.68). The demonstrated feasibility of the implementation at plant scale highlights the opportunities of sensor-based material flow monitoring for the waste management sector and paves the way towards a more circular plastics economy.

AB - Sensor-based material flow monitoring allows for continuously high output qualities, through quality management and process control. The implementation in the waste management sector, however, is inhibited by the heterogeneity of waste and throughput fluctuations. In this study, challenges and possibilities of using different types of sensors in a lightweight packaging waste sorting plant are investigated. Three external sensors have been mounted on different positions in an Austrian sorting plant: one Light Detection and Ranging (LiDAR) sensor for monitoring the volume flow and two near-infrared (NIR) sensors for measuring the pixel-based material composition and occupation density. Additionally, the data of an existing sensor-based sorter (SBS) were evaluated. To predict the newly introduced parameter material-specific occupation density (MSOD) of multi-coloured polyethylene terephthalate (PET) preconcentrate, different machine learning models were evaluated. The results indicate that using SBS data for both monitoring of throughput fluctuations caused by different bag opener settings as well as monitoring the material composition is feasible, if the pre-set teach-in is suitable. The ridge regression model based on SBS was comparable (RMSE = 3.59 px%, R² = 0.57) to the one based on NIR and LiDAR (RMSE = 3.1 px%, R² = 0.68). The demonstrated feasibility of the implementation at plant scale highlights the opportunities of sensor-based material flow monitoring for the waste management sector and paves the way towards a more circular plastics economy.

KW - Sensor-based material flow monitoring

KW - leightweight packaging waste

KW - plastic

KW - sorting plant

KW - machine learning

KW - prediction model

KW - near-infrared

KW - LiDAR

KW - LiDAR

KW - lightweight packaging waste

KW - machine learning

KW - near-infrared

KW - plastic

KW - prediction model

KW - Sensor-based material flow monitoring

KW - sorting plant

UR - http://www.scopus.com/inward/record.url?scp=85192156881&partnerID=8YFLogxK

U2 - 10.1177/0734242X241237184

DO - 10.1177/0734242X241237184

M3 - Article

VL - 42.2024

SP - 747

EP - 758

JO - Waste management & research

JF - Waste management & research

SN - 0734-242X

IS - 9

ER -