Rapid detection of copper ore grade based on visible-infrared spectroscopy and TSVD-IVTELM. (15th November 2022)

• Record Type: Journal Article
• Authors: Xie, Hongfei; Mao, Zhizhong; Xiao, Dong; Liu, Jingyi
• Is Part Of: Measurement. Volume 203(2022)
• Publication Date: 2022-11-15
• Subjects: Ore grade -- Extreme learning machine -- Visible-infrared spectroscopy -- Truncated singular value; Weights and measures -- Periodicals; Measurement -- Periodicals; Measurement; Weights and measures; Periodicals; 530.8
• DOI: 10.1016/j.measurement.2022.112003
• Languages: English
• ISSNs: 0263-2241

Abstract

Highlights: A method for rapid detection of copper ore grade is proposed. The method is based on visible-infrared spectroscopy and machine learning. Spectral data is non-linearly detected by graphical diagnosis and numerical calculation. After comparative test, TSVD-IVTELM has the highest accuracy. Abstract: The rapidity of ore grade identification is key to speeding up the beneficiation process in the mining process. Traditional ore grade detection mostly relies on chemical methods. Although these methods have high accuracy, they take a long time, and the cost of detection has always been high. Therefore, this paper proposes a detection method for ore grade using visible-infrared spectroscopy and an incremental two hidden layer extreme learning machine with variable hidden layer nodes based on the truncated singular value decomposition (TSVD-IVTELM) algorithm. Firstly, the spectral data of each sample are obtained by spectrometer. Then, Monte Carlo cross-validation is used to eliminate abnormal samples, and partial least squares regression is used to extract the latent variables of the spectral data to reduce the data dimension. Finally, TSVD-IVTELM is used for regression analysis. TSVD-IVTELM is proved to have the smallest root mean square error and best fitting performance after comparison experiments.