Information and communication and chemical technologies

Vol. 1 No. 30 (2026) / 2026-03-30 / Number of views: 151

PREDICTIVE MODEL OF LABORATORY CONTROL OF CHEMICAL COMPOSITION OF MINING AND METALLURGICAL PRODUCTS

Authors

Nazarbayev Intellectual School
https://orcid.org/0009-0008-1214-0867
K.Kulazhanov Kazakh University of Technology and Business
https://orcid.org/0009-0007-1870-040X
K.Kulazhanov Kazakh University of Technology and Business
https://orcid.org/0009-0006-3517-1389

Keywords

laboratory control, chemical composition, ferrochrome, machine learning, ensemble model, predictive analytics, digital twin, metallurgical production

Link to DOI:

https://doi.org/10.58805/kazutb.v.1.30-1476

How to quote

Akisheva, L.K., et al. “PREDICTIVE MODEL OF LABORATORY CONTROL OF CHEMICAL COMPOSITION OF MINING AND METALLURGICAL PRODUCTS”. Vestnik KazUTB, vol. 1, no. 30, Mar. 2026, doi:10.58805/kazutb.v.1.30-1476.

Abstract

The relevance of the research is determined by the need to move from laboratory control to predictive quality management of metallurgical products. The delay in laboratory analysis, in practice, increases the risk of product release outside of regulatory requirements and reduces the efficiency of technological solutions.

The paper develops the architecture of an intelligent laboratory chemical composition control system based on the integration of LIMS, SCADA/MES data and real data obtained from open sources. Ensemble models (Gradient Boosting, ANN, LSTM, Stacking) have been implemented to predict the mass fractions of chemical elements.

According to the results of the study, quantitative indicators were obtained: R2 for Cr — 0.94, RMSE — 0.72%, MAE — 0.48%, which is more than two times lower than the permissible standard deviation (± 1.0%). For the elements P and S, the absolute error does not exceed 0.01%. The ensemble model maintains forecasting stability with varying input parameters (±3-5%) and does not reveal a statistically significant bias in estimates, which is confirmed by the Bias value close to zero. Qualitative indicators indicate fairly stable forecasts, interpretability and technological validity of the identified dependencies. The developed system ensures the transition from laboratory to proactive quality management, reducing decision-making time from 2-4 hours to several minutes, reducing the risk of production outside the standards by 15-25%. The practical significance of the research lies in reducing decision-making time, reducing the risk of marriage by 15-25%, and the possibility of integration into the company's digital twin.

Further research is aimed at introducing online learning, expanding to other ferroalloys, and integrating with physico-chemical process models.