Computational Simulation and Prediction on Electrical Conductivity of Oxide-Based Melts by Big Data Mining

Area of Science:

• Materials Science
• Metallurgical Engineering
• Computational Chemistry

Background:

• Electrical conductivity is a fundamental property of oxide melts, crucial for materials and metallurgical industries.
• Traditional experimental measurements for molten slag conductivity are labor-intensive and time-consuming.
• Data-driven decision-making offers a powerful alternative to traditional expert experience in scientific research.

Purpose of the Study:

• To develop and validate a computational simulation approach for predicting the electrical conductivity of molten slags.
• To leverage big data mining techniques for enhanced understanding and prediction of slag conductivity.
• To identify the primary chemical components influencing slag electrical conductivity.

Main Methods:

• Application of big data mining methodologies for computational simulation.
• Development and comparison of predictive models, including the Gradient Boosting Decision Tree (GBDT).
• Experimental validation of the predictive models and analysis of influencing factors.

Main Results:

• The Gradient Boosting Decision Tree (GBDT) model demonstrated superior predictive performance with 90% accuracy and 88% sensitivity.
• The GBDT model's robustness confirmed the reliability of the prediction outcomes.
• Slag conductivity is significantly influenced by TiO₂, FeO, SiO₂, and CaO concentrations.

Conclusions:

• A data-driven approach using GBDT is effective for predicting molten slag electrical conductivity.
• TiO₂ and FeO positively correlate with electrical conductivity, while SiO₂ and CaO exhibit negative correlations.
• The findings provide a foundation for optimizing slag properties through computational methods in industrial applications.