Optimization of the scrap flow of a sensor-based sorting system through a vibratory feeding unit

Abstract

Primary aluminum production is harmful for both humans and the environment. Therefore, advanced recycling technologies are important to make the aluminum supply more circular and less damaging. The Xorter is one such advanced recycling technology from the company Myne.
In this process particles are scanned to identify their alloy type and are ejected from a conveyor belt into different buckets based on this type. For the particles to be identified and ejected one by one, it is important that they are singulated properly so that there is no clustering on the belt. This happens with a vibratory feeder followed by singulation chutes. However, the problem at hand is that the singulation is not efficient enough yet because of fluctuations in the particle rate and poor spreading before the particles enter the chutes.
This research investigates, through modeling, how to optimize the scrap flow by focusing on the vibratory feeder. First, different ways to create a buffer by adding designs on the feeder were tested. It was found that the particle sizes are too diverse to create a predictable buffer that lets only a certain amount of particles go through at a time. Another finding is that imple-menting 7° bumps within narrowings can reduce the fluctuations but is probably only suitable for material flows with bigger and more homogeneous particles than the one the Xorter has to deal with.
Then, two different designs for spreading out the particles over the width of the feeder were tested, and it was found that triangular splitters in the middle of each side of the feeder can help with improving the spreading.
Thereafter, multiple ways to feed the particles onto the vibratory feeder were modeled. Currently, the particle are fed to the feeder with a 90° angle but the clustering of particles at the end of the feeder will decrease when this is changed to the transport direction. Moreover, this would increase the distribution between both sides of the feeder. The optimal belt speed is 1m/s and a sloped plate under the belt could decrease clustering.
The effect of the incoming particle rate was studied as well. A higher average rate leads to significantly more clustering and is therefore not recommended.
Lastly, the connection between the indicators used (spreading and clustering at the end of the feeder) and the singulation rate after the chutes was examined. This revealed that these indicators do not directly represent the singulation rate. Thus, future research should not rely on simplified indicators but include the singulation chutes to make sure that the results are the most accurate.