(中南大學 冶金科學與工程學院， 長沙 410083)
摘 要: 采用BP神經(jīng)網(wǎng)絡對鋁電解NiFe2O4基金屬陶瓷惰性陽極的電解腐蝕過程進行了系統(tǒng)辨識。建立了以Al2O3質量濃度、 電解溫度、 分子比、 面積比和電流密度為輸入，腐蝕率為輸出的網(wǎng)絡模型。 在材料的設計中， 采用了GA-BP優(yōu)化方法， BP網(wǎng)絡參與GA迭代計算時對個體的評價。 應用結果表明， NiFe2O4基金屬陶瓷惰性陽極的電解腐蝕率預測結果與實測值吻合；優(yōu)化設計的結果與實驗值很接近。
關鍵字: 鋁電解； 惰性陽極； 腐蝕； 人工神經(jīng)網(wǎng)絡； 遺傳算法
anodes based on GA-BP hybrid neural net work
（ School of Metallurgical Science and Engineering，
Central South University， Changsha 410083, China）
Abstract: The corrosion processes of 5%Ni-NiFe2O4 inert anodes were recognized by back propagation neural net works and the prediction model was presented. The structures of neural net work include four input nodes, alumina concentration, bath temperature, cryolitic ratio, and area ratio of cathode to anode, current density, and one output node, corrosion rate. The hybrid neural network, genetic algorithms and back propagation neural networks, were applied when optimizing the design of the trial parameters. Some trial strategies were deduced by the hybrid model. The application and experimental results shows that, the neural prediction values of the corrosion rate of NiFe2O4 inert anodes fit in with the trial values, and the hybrid neural network model has guidance signification for material design.
Key words: aluminum electrolysis； inert anode； corrosion； artificial neural network； genetic algorithms
