Abstract
A Neural Network (NN) based on back-propagation algorithm is proposed to predict the coke consumption in a melting furnace, and to control the coke charge automatically. This new-type melting furnace with pulverized dried sludge injection has been developed to enhance the performance of the conventional coke bed melting furnace. Numerous operating conditions and measured data of bench-scale experiments are used as learning data to determine the network architecture and parameters. It should be noted that there is a significant drop in the computing time to implement the network model and perform simulation using the network model compared with the mathematical model. The results predicted by the network model are in reasonable agreement with the actual measurement data and the results by the mathematical model, suggesting the effectiveness and applicability of this model to predict the performance of melting furnace and to control the coke charge automatically. In addition, the network model is utilized to investigate the influence of such operating conditions as the sludge charge rate on performance of the bench-scale melting furnace.
