A novel approach for sensor based sorting performance determination


MINERALS ENGINEERING, vol.146, 2020 (SCI-Expanded) identifier identifier

  • Publication Type: Article / Article
  • Volume: 146
  • Publication Date: 2020
  • Doi Number: 10.1016/j.mineng.2019.106130
  • Journal Indexes: Science Citation Index Expanded (SCI-EXPANDED), Scopus, Academic Search Premier, Aerospace Database, CAB Abstracts, Chemical Abstracts Core, Communication Abstracts, Compendex, INSPEC, Metadex, Veterinary Science Database, Civil Engineering Abstracts
  • Keywords: Borates, Ore sorting, NIR, Process efficiency, ROC, SE (%), Partition curve, Process simulation, COPPER, SEPARATION, MODEL, BULK
  • Hacettepe University Affiliated: Yes


Sensor-based sorters execute an "accept-reject" task to selectively remove the target particles from a material stream. Depending on the intended use, defining the separation performance of a Sensor-based sorting (SBS) operation has critical importance on evaluation of the output product qualities. SBS systems are dominated by line-scan sensors, within which image acquisition via machine vision is practicable as the material is sensed during transportation on a belt or during free fall. Current favorable tools for performance evaluation use numerous parameters such as particle size, throughput, mass % to be rejected, accurate identification of the particles, and even particle shape to determine efficiency, performance, and accuracy of the sorting process in mining/mineral processing industry. However, performance of the operation is also affected by mechanical and physical factors along with the correct identification of materials. Hence, it is necessary to ensure and monitor that all accurately identified particles are reported to the correct product stream, either accept or reject. The aim of this study is to investigate the performance of a real-time Near-infrared (NIR)-sorter with a novel perspective, and compare the results with favorable and known previous approaches. Within the experimental studies, accept and reject products of an industrial scale NIR-sorter used for borate sorting were collected, and sieved to three distinct size fractions. Images of individual particles were acquired with and without optical filters, and boron contents were analyzed, respectively. Resulting data was used to determine the material specific partition coefficients of the sorting operation for different particle size classes. Combined overall and actual separations reported fairly average figures of 0.105 Ep (Ecart probability) and 0.730 RV50 (cut-reflectance value) for the separation. Evaluation of the overall findings showed that presented approach execute reliable performance figures applicable to ore sorting applications.