Characterizing and Controlling Abnormal Periodic Mold Level Fluctuations in a Commercial Slab Continuous Caster Using Big Data

Abstract

The stable control of mold level is a key link in the production of high-quality continuous casting slabs. Periodic mold level fluctuation (PMLF) is common during the continuous casting process, and the abnormal PMLF has significant harmful effects on surface quality of slab. This article proposed an analysis and control method for abnormal PMLF. First, the finite impulse response (FIR) filter and fast Fourier transform (FFT) were used to remove noise interference in PMLF data and highlight the fluctuation characteristics of PMLM. Then, considering that uneven solidification has a significant impact on abnormal PMLF, the influence of chemical composition on the equilibrium Fe-C pseudo-binary diagram was calculated by Thermo-Calc software. Furthermore, roller diameter, roller spacing, casting speed, and chemical composition were chosen as the prediction indicator to predict the quality of PMLF. Random forest (RF) model shows good performance in predicting PMLF; the prediction accuracy of RF model is 92.76 pct, which is 21.39 pct higher than that of GA-BP model. Finally, the Feedforward fuzzy PID (F2FPID) controller designed in this article was used to eliminate abnormal PMLF. The average range of mold level fluctuation under the PID controller is ± 6.8 mm, while under the F2FPID controller, the average range of mold level fluctuation is ± 1.1 mm. And the F2FPID controller owns a lower overshoot of 0.48 pct and an adjusting time of 1.52 seconds, which are 94.8 pct and 59.5 pct, respectively, lower than those of the PID controller.