基于RBF神经网络的添加剂改性炭材料高温黏结剂的性能预报模型

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

在炭材料黏结剂添加剂改性实验数据的基础上,将神经网络方法用于研究添加剂配方和热处理温度对黏结强度的影响关系,建立了添加剂改性炭材料黏结剂的RBF(Radial Basis Function径向基函数)神经网络性能预报模型,并与BP(Back-Propagation逆传播)人工神经网络进行了预报精度和训练过程比较。结果表明:上述两种模型对于黏结强度的预报平均相对误差分别为0.0127和0.0600,且BP人工神经网络易陷入局部最小。因此,RBF神经网络模型的预报能力较好,得出了具有较精确黏结性能的添加剂配方和热处理数据。可望在炭材料黏结剂改性中的多变量、非线性体系中提高实验工作效率,为炭材料黏结剂提供一条有应用前景的理论设计途径。

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 0.0127 and 0.0600 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.