End-point carbon content correction in converter steelmaking based on the Kalman filter and CBR model

Abstract

End-point carbon content at converter is one of significant indicators of end point control. While the bomb-dropping measurement technology can realise quick and effective measurement of end-point carbon content, it is difficult to achieve accurate end-point control due to its lower detection accuracy compared to other methods. Herein, a method was established for correcting the bomb-dropping measurement of end-point carbon content in converter steelmaking to improve its measurement accuracy, thus further achieve cost reduction and efficiency enhancement in converter production. Historical production data and Case-based Reasoning (CBR) model were adopted to establish the prediction model of end-point carbon content and Kalman filtering (KF) was used to fuse CBR model prediction and bomb-dropping measurement to get more accurate end-point carbon content of converter. Through data analysis and on-site tracking and sampling, the validity of correction was thus verified. Experiments showed that: the optimal ratio of the size of training set to test set of CBR prediction model was 95:5 and the optimal Weighting Number was 9; CBR model's average prediction error was 0.019%, bomb-dropping measurement's average error was 0.029% and the average error after KF fusion was 0.014%. The carbon content accuracy was improved by KF fusion by 51.7% compared with that of the original bomb-dropping measurement. Corrected end-point carbon content presented a more typical normal distribution. This method could reduce the measurement error in bomb-dropping.