神经网络低温铝电解电流效率预报模型

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Ｎａ３ＡｌＦ６－ＡｌＦ３－ＣａＦ３－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

PREDICTION MODEL FOR CURRENT EFFLCLENCY IN LOW TEMPERATURE ALUMINIUM ELECTROLYSIS BASED ON NEURAL NETWORKS

The prediction model for current efficiency (CE) of low temperature aluminium electrolysis(LTAE) based on artificial neural networks(ANN) has been investigated in this paper. An experimental sample for the study was taken from experiments of LTAE with low molar ratio eleCtrolyte of Na3 AIF6-AIF3-CaF2--MgF2-LiF-Al2O3 system.Through training ANN With a learning sample chosen at random from experimental sample, the relationship model was developed between CE and technological parameters of LTAE, including bath composition, electrolytic temperature, cathodic current density and polar distance. And the model precision was tested by the remainder of the experimental sample. The testing and predicting results showed that this model possessed high precision and resulted in good predicting effect. ANN as a new cooperating and predicting technology provides new approach to the further studies of LTAE.