End Temperature Prediction of Molten Steel in RH Based on Case-based Reasoning with Optimized Case Base

Data redundancy issue of case-based reasoning (CBR) method can negatively influence, calculation efficiency and prediction accuracy as the production data accumulate when predicting end temperature in Ruhrstahl Heraeus (RH) refining. A case base optimization method was proposed here to resolve the issue. The correlation between different cases in the original case base was analyzed by using similarity, and case base sets were designed in three different principles. Same testing data were used to examine all the cases in the case base set and the optimized case base was obtained via integrated comparison. Results from production data indicated that the case base set with minimum similarity provided optimal case base. Not only was the calculation efficiency enhanced, but the prediction accuracy improved. The research result has practical value to the application of CBR in RH refining in steelmaking industry.