基于BP神经网络的熔盐电导率预测研究  

Electrical Conductivity Rate Prediction of Molten Salt based on BP Neural Network

作　　者：向燕[1] 张小联[1] 郑鑫 

机构地区：[1]赣南师范学院,江西赣州341000 [2]赣州飞腾轻合金有限公司,江西赣州341000

出　　处：《中国钨业》2014年第3期34-37,共4页China Tungsten Industry

基　　金：国家自然科学基金项目(51064001)

摘　　要：NdF3-LiF-Nd2O3体系熔盐电导率是稀土熔盐电解的基础参数，由于高温环境使其在电解过程中的变化规律难以获得。研究针对试验结果所取得的样本进行训练，通过BP神经网络预测了NdF3-LiF-Nd2O3体系熔盐电导率，并分析了温度、LiF浓度和Nd2O3浓度对熔盐电导率的影响。研究结果表明，预测值处在1.8256~3.1197 S/cm之间，与实验值误差在3%左右，同时，预测值与实际值的变化趋势基本一致。研究表明BP神经网络对熔盐电导率的预测能够满足熔盐体系电导率研究的要求。The basic parameter of rare earth electrolys is the electrical conductivity of NdF3-LiF-Nd2O3 molten salts. However, there is difficult to obtain the changes of the process of electrolysis because of the high temperature environment. With the experimental results obtained in the research on the training samples, to predict the conductivity of NdF3-LiF-Nd2O3 molten salts by the BP neural network, and analyze the influence of temperature, LiF and Nd2O3 on the electrical conductivity of molten salts. The research results show that the predicted values in 1.825 6-3.119 7 S·cm-1, and the experimental value of error is about 3%, predictive value and the actual value of the change tendency is consistent. Researches show that prediction of conductivity of molten salts by BP neural network can meet the requirements of electrical conductivity of molten salts system.

关 键 词：BP神经网络 MATLAB电导率 预测模型 NdF3-LiF—Nd2O3 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程] TF845[自动化与计算机技术—控制科学与工程]