Neural Network Prediction of Conversion Rate of TbFe2 Alloy Prepared by Reduction-Diffusion Process

摘要:

针对还原扩散法制备TbFe2合金的主要实验参数：反应温度、保温时间、Ca的加入量及Fe的粒度，建立BP神经网络，进行仿真，预测TbFe2合金的转化率。以44组实验数据作为训练样本，进行了网络设计。通过测试及对网络的性能分析，证明了该网络能够准确预测不同实验参数下TbFe2合金的转化率，并具有良好的性能。该网络的设计可以缩短实验周期，节约实验成本，并对反应的机理及工艺研究有一定的价值。

Abstract:

A BP neural network was established based on the following main experiment parameters of producing TbFe2 alloy by reduction-diffusion process: reaction temperature, holding time, quantity of Ca and particle size of Fe. A simulation was conducted, and the rate of conversion of TbFe2 alloy was predicted. The neural network was simulated and tested by 44 groups of experimental data. It can be concluded that the neural network has good performance to predict the rate of conversion of TbFe2 alloy. The design and the application of this neural network can help to shorten the periodic time of experiments, lower the experimental cost, and optimize the preparation processes.