Premium
Monitoring the blast furnace process using neural networks and knowledge‐based system
Author(s) -
Zuo Guangqing,
Björkman Bo
Publication year - 2001
Publication title -
steel research
Language(s) - English
Resource type - Journals
eISSN - 1869-344X
pISSN - 0177-4832
DOI - 10.1002/srin.200100094
Subject(s) - thermal state , thermal , artificial neural network , process (computing) , channelling , fuzzy logic , blast furnace , engineering , computer science , materials science , artificial intelligence , metallurgy , chemistry , physics , thermodynamics , operating system , ion , organic chemistry
To assist the daily control of blast furnace operation, models for diagnosing the irregular process status and for predicting the cool furnace thermal state have been developed. The diagnostic model consists of two sub‐models. One is for evaluation of the various aspects of the process status based on fuzzy logic. The other is for the detection of the occurrence of channelling in the furnace based on neural networks. Tests using the actual process data have shown that the former sub‐model can promptly detect the existing abnormal process status and give warnings of irregular process statuses, e.g. abnormal permeability of burden, high heat fluxes, etc. The latter sub‐model can successfully extract the characteristic data patterns from a large amount of process data in connection with the occurrence of channelling and detect the existing channelling. For appropriately depicting the furnace thermal state, a thermal index and characteristic patterns of cool furnace thermal state have been derived using fuzzy logic and neural networks. Accordingly, two neural network models were designed for predicting the cool furnace thermal state. Either thermal index or characteristic patterns can be used to represent the cool furnace thermal state. Either model can be used to predict the upcoming cool furnace thermal state.