论文部分内容阅读
本文研究基于神经网络的低温铝电解电流效率预报模型．实验样本采自用Wa3AIF6－AIF3－CaF2－MgF2－Lif－Al2O3体系中低分子比电解质所进行的电解实验．从实验样本中随机抽取学习样本训练网络．建立电流效率与影响它的电解工艺参数包括熔体成分、电解温度、阴极电流密度和极距之间的关系模型，尔后用剩余的实验样本检验模型精度．结果表明：该模型精度高，具有良好的预报效果．神经网络作为一种新颖的拟合预报新技术，为低温铝电解的研究提供了新的途径．
This paper studies the prediction model of current efficiency of low-temperature aluminum electrolysis based on neural network. Experimental samples were taken from the electrolysis experiments using low molecular weight electrolytes in the Wa3AIF6-AIF3-CaF2-MgF2-Lif-Al2O3 system. Learning samples were randomly selected from the experimental samples to train the network. Establish the current efficiency and its impact on the electrolysis process parameters including melt composition, electrolysis temperature, cathode current density and polar distance between the model, and then use the remaining experimental samples to test model accuracy. The results show that the model has high accuracy and good prediction effect. As a novel novel technique of fitting prediction, neural network provides a new way for the research of low temperature aluminum electrolysis.