Hot metal temperature prediction and simulation by fuzzy logic in a blast furnace

This work describes the development and further validation of a model devot ed to blast furnace hot metal temperature forecast, based on Fuzzy logic pr inciples. The model employs as input variables, the control variables of an actual blast furnace: Blast volume, moisture, coal injection, oxygen addit ion, etc. and it yields as a result the hot metal temperature with a foreca st horizon of forty minutes. As far as the variables used to develop the mo del have been obtained from data supplied by an actual blast furnace sensor s, it is necessary to properly analyse and handle such data. Especial atten tion was paid to data temporal correlation, fitting by interpolation the di fferent sampling rates. in the training stage of the model the ANFIS (Adapt ive Neuro-Fuzzy inference System) and the Subtractive Clustering algorithms have been used.