Predicting coal ash fusion temperature with a back-propagation neural network model

Abstract

A novel technique, the back-propagation (BP) neural network, is presented for predicting the ash fusion temperature from ash compositions for some Chinese coals instead of the traditional techniques, such as the ternary equilibrium phase diagrams and regression relationships. In the applications of the BP networks, some modifications to the original BP algorithm are adopted to speed up the BP learning algorithm, and some useful advice is put forward for the choice of some key parameters in the BP model. Compared to the traditional techniques, the BP neural network method is much more convenient and direct, and can always achieve a much better prediction effect.