Original Article

Intelligent control for LF process operations and endpoints based on mixed-algorithm model

State Key Laboratory of Advanced Metallurgy, University of Science and Technology Beijing, Beijing 100083, P.R. China

^* e-mail: sysong@ustb.edu.cn
^** e-mail: lijing@ustb.edu.cn

Received: 15 November 2024
Accepted: 3 March 2025

Abstract

An intelligent control system for LF process operations and endpoint prediction has been developed, leveraging a mixed-algorithm model. This model incorporates historical production data from LF refining processes spanning the years 2021 to 2023, stored within its database. The model’s architecture initiates with the long short-term case-based reasoning (LST-CBR) methodology, enhanced with the coupled reaction calculation module, forming the intelligent mixed-algorithm LF refining mechanism. The CBR module, by drawing parallels between current scenarios and past cases, provides forecasts for refining endpoints alongside recommends operational strategies. Subsequently, these recommendations are enacted in real-time, directing the LF equipment operations via PLC commands, not only give instructions to any part of LF refining systems, but also receive return signals. The prediction results based on 120-ton industrial ladle furnace showed a reasonable operation selection, all alloy addition has an effective control during the LF process. Furthermore, the overall refining duration, measured from the start of the converter tapping process to the conclusion of LF treatment, was shortened from 63 min to 50 min. Additionally, the time required for process heating was curtailed from 40 min to a mere 25 minutes. Finally, the accuracy of the model continually evolved through a self-learning mechanism, benefiting from the constant new massive data addition. Accordingly, the intelligent LF refining model was stable and suitable for large-scale prediction in LF refining endpoint and process operations.