DEVELOPMENT AND RESEARCH OF A SYNTHETIC DATA GENERATOR FOR MODELING TECHNOLOGICAL PARAMETERS OF THE IRON ORE BENEFICIATION PROCESS

The paper presents the results of developing a specialized synthetic data generator for modeling technological parameters of the iron ore beneficiation process.The developed generator implements a modular approach that includes parameter configuration, basic signal generation, and controlled noise addition.Methods for generating basic signals are proposed considering the specifics of various technological parameters through the use of normal distribution for parameters with natural grouping tendencies and uniform distribution for parameters with uniform change patterns.The noise addition system provides four intensity modes with individual settings for each parameter.The experimental study included 486 experiments with different training sample configurations.Analysis of generation quality showed that SNR values are in the range of 25-50 dB, which corresponds to the characteristics of real industrial measurements.A critical dependence of prediction quality on training sample size was established: with sample sizes over 4096 records, consistently high prediction quality is achieved (R 2 = 90-95 %).The developed generator enables the creation of synthetic datasets with controlled noise characteristics and statistical properties that correspond to real technological parameters.Validation mechanisms for generated data are implemented through comprehensive analysis of statistical characteristics, spectral analysis, and evaluation of correlation dependencies.The results confirmed the effectiveness of the developed generator for creating training datasets in the development and testing of control systems for iron ore beneficiation processes, which significantly reduces time and resources during the control system development phase.