Classification of Wrought and Cast Aluminium using Magnetic Induction Spectroscopy and Machine Vision

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

Abstract

Recycled aluminium can reduce the greenhouse gas emissions created and energy required to produce aluminium compared to virgin bauxite ore. Once aluminium is separated from other non-ferrous metals, it is labelled ‘Twitch’ and consists of wrought and cast aluminium. Wrought is removed to avoid contamination from the cast pieces, as contamination undermines the alloy’s sustainability and changes the metals’ properties. In this paper, we demonstrate the use of magnetic induction spectroscopy to classify wrought from cast independently and combined with a machine vision camera on a conveyor. The magnetic induction sensor measures 6 frequencies between a range of 2736 to 59508 Hz. The camera extracts the colour, perimeter, area and offset of the metal piece. The combinations of induction, induction and shape, induction and colour, and colour are tried to determine the best sensor combination. We first show how wrought can be classified with induction only with a 71.21-85.58% recovery and 74.6-83.26% purity. We then show how the combination of induction and the colour of the metal pieces as features can increase the recovery to 71.21-92.56% and the purity to 83.92-88.05%. Classification using colour only obtained an F1 score of 0.598-0.789, whereas induction only had an F1 score of 0.844-0.729. The addition of shape as a feature did not noticeably improve the recovery and purity.
Original languageEnglish
Title of host publication2024 IEEE Sensors Applications Symposium (SAS)
Publication statusAccepted/In press - 14 May 2024

Publication series

NameAdvanced and Intelligent Sensor Applications for a Better Future

Keywords

  • Aluminium
  • Electromagnetic induction
  • Recycling
  • Spectroscopy
  • Waste recovery
  • Magnetic induction spectra
  • Machine vision

Fingerprint

Dive into the research topics of 'Classification of Wrought and Cast Aluminium using Magnetic Induction Spectroscopy and Machine Vision'. Together they form a unique fingerprint.

Cite this