Multi-output support vector regression modeling for multivariate molten iron quality indices in blast furnace ironmaking process
ZHOU Ping
LI Rui-feng
GUO Dong-wei
WANG Hong
CHAI Tian-you
Abstract：Molten iron temperature as well as Si, P, and S contents are the most essential molten iron quality (MIQ) indices in the blast furnace (BF) ironmaking, while difficult to be directly measured online, and large-time delay exists in offline analysis through laboratory sampling. Focusing on this practical challenge, a data-driven multi-output support vector regression (M–SVR) dynamic model is established to estimate the MIQ indices online, with the help of the pro-posed comprehensive evaluation on modeling accuracy and genetic optimization on model parameters. Different from the conventional single output SVR, the M–SVR can calculate multiple classification by one training process, so as to realize multi-output regression modeling of multivariate MIQ indices. With the proposed comprehensive evaluation index on mod-eling accuracy, the modeling performance can be evaluated from the aspects of model estimation trend as well as estimation error. By taking this comprehensive evaluation index as the fitness function, the genetic algorithm (GA) is to find the opti-mal values of the telescopic vector and the penalty factor for the M–SVR model, so that the GA–M–SVR dynamic model with optimal parameters can be obtained. Finally, industrial experiments have been carried out on the 2# blast furnace in an Iron&Steel Group Co. of China, where it has been demonstrated that the GA–M–SVR model produces satisfied modeling and estimating accuracy.
Keywords：blast furnace ironmakingmolten iron quality (MIQ)multi-output support vector regression (M-SVR)comprehensive evaluation of modeling accuracygenetic optimizationdata-driven modeling
Publication Date：2016-01-01
Online Publishing Date：2025-08-15(First online date of this platform, not the publication date of the document)
Pages：8( 727-734 )