人工神经网络在玻璃配方设计中的应用研究

应用人工神经网络技术，采用Neuralworks Predict软件建立BP网络模型，通过对R2O-MO-Al2O3-SiO2系统玻璃组成与热膨胀系数关系实验数据的训练，以期能预测该系统指定组成的玻璃的热膨胀系数?研究结果表明，所建立的神经网络模型能较正确地反映玻璃氧化物组成与其热膨胀系数之间的规律性。模型对给定组成玻璃热膨胀系数的预测值与实际测试值的相对误差在6．4％以内，表明由神经网络技术建立的这一模型能正确反映R2O-MO-Al2O3-SiO2系统玻璃组成与热膨胀系数间的内在规律性。

Application of Artificial Neural Networks to Glass Composition Design

In order to gain the regularity between the glass composition and the thermal expansion coefficient of R_2O-MO-Al_2O_3-SiO_2 glass system,an artificial neural network model has been established by using of the Neuralworks Pre- dict technology.The model has been trained by a serial data from the experiments.It is founded that the thermal expansion coef- ficients based on the model for R_2O-MO-Al_2O_3-SiO_2 glass may fit in with the actual values.With the help of the model,the thermal expansion coefficients of the glasses are predicted.The relative error between the predieted values and the experimental results is less than 6.5%,which shows that the model established by ANN can response the inner regularity between the glass composition and the thermal expansion coefficients of the glass system.