机构地区: 中南大学
出 处: 《稀有金属材料与工程》 2005年第5期726-730,共5页
摘 要: 通过对实验Al-Zn-Mg-Cu-Zr-Ag合金不同温度下(90℃~150℃)时效得到的硬度和导电率数据进行了神经网络建模,发现在目标函数为0.3,隐层节点数为5,学习率为0.15时,系统误差较小.利用所建立的网络模型预测不同时效状态下材料的硬度和导电率值,发现预测数据与实验数据吻合良好(总误差3.5%),为铝合金时效性能预测和控制提供了1条新途径. An artificial neural network model was constructed using the data of hardness and conductivity gained by ageing at various temperatures ranging from 90 degrees C similar to 150 degrees C for Al-Zn-Mg-Cu-Zr-Ag alloy. It is found that the systematical error is small when the value of objective function is 0.3, the number of nodes in the hidden layer is 5 and the learn-rate is 0.15. The prediction results of hardness and conductivity showed good agreement with the experiment data. This model provided a new road for the prediction and control of the ageing properties of aluminum alloys.