Transactions on Machine Intelligence

Transactions on Machine Intelligence

Artificial Intelligence for Matte Grade Prediction in Copper Flash Smelting

Document Type : Original Article

Author
Hossein Ghayoumi Zadeh
Associate Professor, Department of Electrical Engineering, Vali-e-Asr University of Rafsanjan, Rafsanjan, Iran
10.22034/tmi.2025.256841
Abstract
The copper content of matte is a critical performance indicator in flash smelting; however, laboratory assay results are typically unavailable until after tapping. This study evaluated two soft-sensing approaches for estimating matte copper grade at the time of tapping using industrial records comprising 8,653 matte taps, 9,012 slag taps, and 7,032 hourly process observations. The first approach employed ridge autoregression with exogenous inputs derived from process measurements and lagged assay results. The second adapted a stoichiometric oxygen-balance model using recent assay information. Performance was evaluated across five rolling-origin test periods covering 4,533 taps over 162 days, assuming a two-hour assay delay. Ridge autoregression achieved the lowest pooled mean absolute error (MAE) of 1.519 percentage points of copper (95% confidence interval: 1.389–1.626), significantly outperforming persistence, which achieved an MAE of 1.592 percentage points (p = 0.0019). Incorporating assay feedback reduced the mass-balance model error substantially, from 4.694 to 1.599 percentage points. The recorded target generated by the commercial advisory system showed a mean absolute deviation of 1.813 percentage points from the measured matte grade. Overall, the findings demonstrate the value of a compact data-driven predictor and show that incorporating updated assay information can substantially improve mass-balance estimates. Further validation using actual assay release times and strictly forward-looking feed imputation is required to establish real-time operational performance.
Keywords
Copper Flash Smelting
Matte Grade
Ridge Regression
Soft Sensor
Rolling-Origin Validation
[1]      Taskinen, P., Jokilaakso, A., Lindberg, D., & Xia, J. (2020). Modelling copper smelting–the flash smelting plant, process and equipment. Mineral Processing and Extractive Metallurgy, 129(2), 207–220. https://doi.org/10.1080/25726641.2019.1688904
[2]      Zhou, J., Chen, Z., Zhou, P., Yu, J.-P., & Liu, A.-M. (2012). Numerical simulation of flow characteristics in settler of flash furnace. Transactions of Nonferrous Metals Society of China, 22(6), 1517–1525. https://doi.org/10.1016/S1003-6326(11)61350-2
[3]      Jylhä, J. P., & Jokilaakso, A. (2023). CFD-DEM models for matte droplet settling in a flash smelting settler. Heliyon, 9(11), e21570. https://doi.org/10.1016/j.heliyon.2023.e21570
[4]      Gui, W.-H., Wang, L.-Y., Yang, C.-H., Xie, Y.-F., & Peng, X.-B. (2007). Intelligent prediction model of matte grade in copper flash smelting process. Transactions of Nonferrous Metals Society of China, 17(5), 1075–1081. https://doi.org/10.1016/S1003-6326(07)60228-3
[5]      Yang, C.-H., Xie, M., Gui, W.-H., & Peng, X.-B. (2008). A prediction model for matte grade in copper flash smelting process. Information and Control, 37(1), 28–33. https://xk.sia.cn/en/article/id/11
[6]      Liu, J.-H., Gui, W.-H., Xie, Y.-F., & Yang, C.-H. (2014). Dynamic modeling of copper flash smelting process at a smelter in China. Applied Mathematical Modelling, 38(7–8), 2206–2213. https://doi.org/10.1016/j.apm.2013.10.035
[7]      Zhang, J., Li, Z., Wang, E., Yu, B., Li, J., & Ma, J. (2024). A temporal network based on characterizing and extracting time series in copper smelting for predicting matte grade. Sensors, 24(23), 7492. https://doi.org/10.3390/s24237492
[8]      Zou, Z., Zhou, C., & Yang, C. (2025). Knowledge-guided encoder-decoder network for soft sensor of copper matte grade in flash smelting process. Neurocomputing, 656, 131485. https://doi.org/10.1016/j.neucom.2025.131485
[9]      He, M. Q., Yu, G., Yang, C., & Han, L. (2023). Dynamic soft sensor modeling of matte grade in copper oxygen-rich side blow bath smelting process. Measurement, 223, 113792. https://doi.org/10.1016/j.measurement.2023.113792
[10]   Cerqueira, V., Torgo, L., & Mozetič, I. (2020). Evaluating time series forecasting models: An empirical study on performance estimation methods. Machine Learning, 109(11), 1997–2028. https://doi.org/10.1007/s10994-020-05910-7
[11]   Tashman, L. J. (2000). Out-of-sample tests of forecasting accuracy: An analysis and review. International Journal of Forecasting, 16(4), 437–450. https://doi.org/10.1016/S0169-2070(00)00065-0
[12]   Hoerl, A. E., & Kennard, R. W. (1970). Ridge regression: Biased estimation for nonorthogonal problems. Technometrics, 12(1), 55–67. https://doi.org/10.1080/00401706.1970.10488634
[13]   Künsch, H. R. (1989). The jackknife and the bootstrap for general stationary observations. The Annals of Statistics, 17(3), 1217–1241. https://doi.org/10.1214/aos/1176347265
[14]   Diebold, F. X., & Mariano, R. S. (1995). Comparing predictive accuracy. Journal of Business & Economic Statistics, 13(3), 253–263. https://doi.org/10.1080/07350015.1995.10524599
[15]   Newey, W. K., & West, K. D. (1987). A simple, positive semi-definite, heteroskedasticity and autocorrelation consistent covariance matrix. Econometrica, 55(3), 703–708. https://doi.org/10.2307/1913610
[16]   Schmidt, A., Guy, B. M., Montenegro, V., Reuter, M., Charitos, A., Stelter, M., & Richter, A. (2023). Flue dust reactions and sticking mechanisms in a copper flash smelting furnace waste heat boiler: A sampling study. Journal of Sustainable Metallurgy, 9, 848–859. https://doi.org/10.1007/s40831-023-00691-4
Transactions on Machine Intelligence
Volume 8, Issue 4
Autumn 2025
Pages 259-268

  • Receive Date 08 August 2025
  • Revise Date 28 November 2025
  • Accept Date 27 December 2025