Application of BP neural network to the prediction of coal ash melting characteristic temperature

摘要

The characteristic temperature of coal ash melting strongly influences the blast furnace injection process. The coal ash deformation temperature (DT) is determined by its chemical composition, but relationship between the two remains uncertain. In this paper, the traditional linear regression, Factsage calculation, and back-propagation (BP) neural network calculation are used to predict the coal ash deformation temperature. The results show that the melting characteristic temperature of coal ash has a great relationship with the coal ash composition. The linear regression can predict the change trend of coal ash deformation temperature, but the prediction results are not very satisfactory. The calculation results of Factsage show a great deviation from the experimental values. The prediction results of the BP neural network can achieve good accuracy, and the maximum relative average error of prediction results is 6.67%. This also illustrates the feasibility of using the BP network prediction model in predicting coal ash deformation temperature.