Quality traceability of converter steelmaking based on adaptive feature selection and multiple linear regression

Steel production is complicated. A production cycle includes multiple steps, each of which leads to generating a large amount of new data, for production information and experience knowledge as example. Analyzing these data and tracing why they were generated contribute to controlling parameters tuning, so that we could guarantee the quality of steel production. Besides, discovery of abnormal situation helps detect the problem with machine or production flow on time, which is crucial for modifying the models. We focused on locating the key parameters when problem appears during production, and proposed a specific method based on improved adaptive feature selection and multiple linear regression approach to realize process analysis and cause traceability for abnormal production data of the converter steelmaking process. Results of experiment suggest that method proposed in this paper is effective and robust, judging by professional knowledge and technological mechanism.