Pattern analysis of the combustions of various copper concentrate tablets using high-speed microscopy and video-based deep learning

• Main Authors: Goto, Y. (Author), Natsui, S. (Author), Nogami, H. (Author), Takahashi, J.-I (Author)
• Format: Article
• Language: English
• Published: Elsevier Ltd 2023
• Subjects: Classification system; Combustion; Combustion behaviours; Combustion test; Copper concentrates; Copper smelting; Deep learning; Energy utilization; Explainable deep learning; Flash smelting; High speed microscopy; High-sped video; Microscopic videography; Pattern analysis; Probability distributions; Silica; Silicon; Video recording
• Online Access: View Fulltext in Publisher

 The combustion behavior of the complex, ever-changing Cu concentrate with SiO2 flux in a flash smelting shaft should be comprehensively understood to improve the efficiency and energy consumption of smelting. To characterize the combustion behavior of each sample, combustion studies involving high-speed digital microscopy and thermal measurements were performed using numerous small Cu concentrate tablets. Generally, two temperature ranges of heating retardation were observed during combustion, at approximately 800–1000 and 1150–1200 °C. A novel video-based Cu concentrate classification system was used to successfully recognize the different combustion patterns of tablets with Cu concentrate-SiO2 mixtures under oxidation gas. This classification system also enabled the calculation of the chemical composition of the concentrate by transforming the network output into a probability distribution. The algorithm based on deep learning employed in this study could learn the combustion behaviors of SiO2-containing Cu concentrates using time-series images extracted from video data. © 2023 The Author(s)