Analysis of the Steelmaking Process via Data Mining and Pearson Correlation.

Area of Science:

• Metallurgical Engineering
• Materials Science
• Process Optimization

Background:

• Continuous improvement in steelmaking is vital.
• Controlling calcium (Ca) and sulfur (S) content is essential for inclusion modification in Ca-treated Al-killed steel.

Purpose of the Study:

• To identify key process variables influencing Ca and S content at the end of ladle furnace refining.
• To apply machine learning for predicting successful steelmaking heats.

Main Methods:

• Utilized a decision tree classifier, a machine learning technique.
• Employed Pearson correlation to link process variables to the root node attribute.
• Analyzed the sulfur distribution coefficient to differentiate satisfactory from unsatisfactory heats.

Main Results:

• The sulfur distribution coefficient at the end of refining was the primary discriminator for heat quality.
• High correlation was found between the sulfur distribution coefficient and end-of-process S content in steel and slag, and Si content.
• Secondary correlations involved Si content and slag basicity with S content.

Conclusions:

• Initial steel and slag conditions at the start of refining are critical.
• Efficient sulfur removal during the refining process is paramount for achieving desired Ca and S levels.
• Optimizing these factors leads to successful Ca-treated Al-killed steel production.