ISSN 2594-357X
Título
Autoria
DOI
Downloads
Resumo
The prediction of productivity represents an important resource in order to anticipate losses, increasing the performance of the sinter machine. In this context, a neural network model which relates the main operating parameters and the sintering productivity was developed. The kind of neural network chosen was multi-layer perceptron. The best configuration was obtained with one hidden layer and nine neurons, and the correlation coefficient obtained between predicted and actual productivity was 0.77. Furthermore, the neural network showed better predictive ability than the multiple linear regression technique. The model was applied to Sinter Machine#3 at Usiminas - Ipatinga Plant.
The prediction of productivity represents an important resource in order to anticipate losses, increasing the performance of the sinter machine. In this context, a neural network model which relates the main operating parameters and the sintering productivity was developed. The kind of neural network chosen was multi-layer perceptron. The best configuration was obtained with one hidden layer and nine neurons, and the correlation coefficient obtained between predicted and actual productivity was 0.77. Furthermore, the neural network showed better predictive ability than the multiple linear regression technique. The model was applied to Sinter Machine#3 at Usiminas - Ipatinga Plant.
Palavras-chave
Sinter plant; Neural network; Productivity.
Sinter plant; Neural network; Productivity.
Como citar
Silva, Thiago Pinto;
Silva, Alexandre Medeiros da;
Resende, Bruno Alves;
Silva., Braulio Viegas da.
PREDICTION OF SINTER PLANT PRODUCTIVITY BY
NEURAL NETWORK
,
p. 764-776.
In: 42º Seminário de Redução de Minério de Ferro e Matérias-primas / 13º Seminário Brasileiro de Minério de Ferro / 6th International Congress on the Science and Technology of Ironmaking,
Rio de Jabeiro,
2012.
ISSN: 2594-357X
, DOI 10.5151/2594-357X-22152
