A radial basis function neural network based model to predict properties of high-temperature-binder additive modified carbon materials

A radial basis function(RBF)neural network predictive model was applied to the experimental data on additive of high temperature binder to join carbon materials. The predictive accuracy and the training process using a RBF neural network were compared with those of a back-propagation (BP) neural network. Results showed that the average relative errors of two such models were 0.012 7 and 0.060 0 for RBF and BP model respectively and BP neural network was easy to fall into a local minimum. Therefore, the RBF neural network predictive model was better than the BP model. It is expected that the RBF neural network can be used in multivariable, nonlinear system to quickly optimize experimental parameters, such as the optimum amount of binder that should be added to the carbon material to achieve a particular property.