Research on processing methods of special fluctuation data in blast furnace big data analysis

Abstract

During the digital and intelligent transformation of the steel industry, big data analysis in blast furnace ironmaking has become a core component for optimizing technological processes and enhancing production efficiency. However, the blast furnace ironmaking process is characterized by multi-variable coupling, strong nonlinear dynamics, and other significant smelting features, leading to complex issues in data collection, such as sample missing, abnormal fluctuations, periodic disturbances, mismatched sampling frequencies, and heterogeneous data types, which traditional data processing techniques struggle to address with precision. In this paper, for the characteristics of blast furnace data, it is proposed to combine the deletion method with Lagrange interpolation method according to the proportion of missing data to make up for the missing data. According to the furnace conditions, the global 3σ rule and the “global + local” 3σ rule are reasonably selected for the processing of occasional perturbation data. Elimination of cyclic disturbances such as hot blast furnace changeovers or blowbacks by specific methodological treatments. The first-order hysteresis filtering method is used to process the data whose change characteristics are masked by fluctuations, and the interpolation method and hierarchical division coding are used to realize the conversion of discrete and continuous data types. The constructed data processing system significantly improves data integrity, consistency, and usability, providing high-quality data foundations for process optimization modeling, abnormal condition warning, and intelligent decision-making systems in blast furnaces, and facilitating key technological breakthroughs in the intelligent upgrading of the steel industry.