Judgment model for copper converter blowing stages based on weighted fusion strategy

In response to the technical challenge of accurately determining multiple stages of copper matte blowing, this study proposes an intelligent judgment method for converter blowing stages based on a weighted fusion strategy. A flame image recognition model and a flue gas judgment model based on the Stacking ensemble algorithm are developed, which integrate flame image and flue gas concentration features to improve the accuracy of the intelligent judgment model for blowing stages. The image recognition model selects RF, XGBoost, and GRNN as base-learner, where RF as the meta-learner, and 5-fold cross-validation is used for training and optimization. The accuracy of recognizing different blowing stages exceeds 93%. Building on this, a weighted fusion strategy is applied to integrate the flame image and flue gas judgment models for comprehensive assessment of the blowing stages, enhancing its accuracy to over 95%. Experimental results demonstrate that the weighted fusion model effectively overcomes the limitations of single-model approaches, significantly improving its capability and accuracy in assessing multiple stages of converter blowing, exhibiting good practical value.