Comprehensive methodology for creating enhanced datasets for modelling the process of magnetic separation of iron ore

The study presents an innovative approach to creating extended datasets for modelling magnetic separation of iron ore, which is crucial for enhancing efficiency and automating enrichment processes in the mining industry. The aim of the research was to develop a methodology for creating extended datasets for modelling magnetic separation of iron ore that takes into account the specifics of Ukrainian deposits and allows for the generation of representative data in conditions of limited real production data by integrating physical modelling with machine learning methods. Research methods: modelling using mathematical learning, simulation based on physical processes, statistical analysis. The study examined the use of the USIM PAC simulator for modelling the iron ore enrichment system and adapting data for magnetic enrichment, ensuring the accuracy of modelling technological enrichment processes. The simulator was used to obtain a dataset from physical modelling of part of the enrichment process based on data from the Valyavkinske deposit. Primary modelling of the dataset was analysed, including statistical characteristics, distribution shape, and normality tests to identify fields requiring correction. Based on the analysis results, specific requirements for data distribution in the new dataset to be formed for further use were established. In accordance with these requirements, several mathematical models were implemented to reproduce the specified criteria and parameters. For each data field, the best model was carefully selected, and the dataset was corrected based on its data to bring the distribution as close as possible to the desired one. Comprehensive validation of the resulting corrected data was conducted, emphasising the preservation of the physical validity of the data and their correspondence to real enrichment processes. A detailed analysis of the corrected data was performed, as well as the statistical characteristics of the resulting dataset, confirming the effectiveness of the developed comprehensive methodology for modelling and adapting data for magnetic enrichment of iron ore. The methodology holds practical value due to its innovative approach to creating extended datasets for modelling magnetic separation of iron ore, enhancing the efficiency and automation of enrichment processes while considering the specifics of deposits and generating representative data in conditions of limited real data