神经网络预测还原扩散法制备DyFe2合金转化率的研究

Neural Network Prediction of Transformation Efficiency of DyFe2 Alloy Prepared by Reduction-Diffusion Process

摘要:

针对还原扩散法制备DyFe2合金中的主要实验参数：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DyFe2合金的转化率。以44组实验数据作为训练样本，进行网络设计，并对网络进行了测试。证明该网络能够预测不同实验参数下DyFe2合金的转化率，且具有良好的性能。该网络的设计可以缩短实验周期，降低实验成本，并有利于工艺优化。

Abstract:

Based on the main experiment parameters of DyFe2 alloy preparation by reduction-diffusion process: reaction temperature, holding time, added quantity of Ca and particle size of Fe, the BP neural network was established and used to predicate the transformation efficiency of DyFe2 alloy. The neural network was simulated by 44 groups of experimental data and was tested. It has been proved that the neural network has good performance to predict the transformation efficiency of DyFe2 alloy. This design of neural network is able to shorten the time of experiment, reduce the experiment cost, and optimize the preparation processes.