ISSN 2594-5327
52º Congresso anual — Vol. 52 , num. 1 (1997)
Título
Autoria
DOI
Downloads
Resumo
Stabilisation of hot metal temperature is the most important objective of blast furnace control actions, both from the point of safety of the plant, and the quality of the products. The availability of real time information concerning the future trends of the hot metal temperature is important to optimise the control action. In the present paper, the activity concerning the development of a hot metal temperature predictor in blast furnace is described. The problem is mainly approached in terms of auto-regressive analysis, reaching accuracy over the 90% in the temperature forecasting in the short-medium time (1-3 hours). The effect of various parameters i.e. the time windows length, the data filtering and the size of the training data sets are discussed. Multi-regressive approaches were tested too, adding in input of the Radial Basis Function structures the main process parameters as the tuyere conditions and the top gas analysis and characteristics.
Stabilisation of hot metal temperature is the most important objective of blast furnace control actions, both from the point of safety of the plant, and the quality of the products. The availability of real time information concerning the future trends of the hot metal temperature is important to optimise the control action. In the present paper, the activity concerning the development of a hot metal temperature predictor in blast furnace is described. The problem is mainly approached in terms of auto-regressive analysis, reaching accuracy over the 90% in the temperature forecasting in the short-medium time (1-3 hours). The effect of various parameters i.e. the time windows length, the data filtering and the size of the training data sets are discussed. Multi-regressive approaches were tested too, adding in input of the Radial Basis Function structures the main process parameters as the tuyere conditions and the top gas analysis and characteristics.
Palavras-chave
blast furnace, hot metal temperature, neural networks, temperature prediction, process control
blast furnace, hot metal temperature, neural networks, temperature prediction, process control
Como citar
Falzetti, Marco; Bellomo, Pietro.
IMPROVING THE BLAST FURNACE CONTROL CAPABILITIES BY MEANS OF A HOT METAL TEMPERATURE ARTIFICIAL NEURAL NETWORKS PREDICTOR,
p. 1984-1994.
In: 52º Congresso anual,
São Paulo, Brasil,
1997.
ISSN: 2594-5327, DOI 10.5151/2594-5327-C00126-1984-1994
