Modeling of temperature change of liquid steel in BOF by neural network

Abstract

To model temperature change of the liquid steel in BOF's process during tapping and adding some additives using neural network. Extent of influences of the network parameters and process variables are studeid. The actual measured data from a steel plant are used as a reference. The study shows that the neural network is capable of predicting the change of the liquid steel temperature during BOF operation and transferring of the liquid steel to the ladles. The forecast temperatures agree with the measured values. It was found that the optimized architecture of the neural network consists of 11 inputs, 4 hidden neurons and 1 output with learning rate and momentum of 0.01 and 0.5 respectively. The discrepancies of the forecast model toe the real values were found to be +- 7 ํC. A model based on thermodynamic and heat balance was also developed and was found to correlate well with the forecast from the neural network. Both models illustrate linear dependency of the temperature on the metallurgical and process parameters. The main factor which causes the temperature drop of liquid steel is tapping time while the steel weight least affects the temperature drop.