【摘要】 建立了基于尺寸效应的Cu-Ag合金抗拉强度反向传播神经网络和粒子群优化神经网络模型,将神经网络模型应用于Cu-Ag合金微细丝抗拉强度的预测。进行不同线径和银含量(质量分数)的Cu-Ag合金微细丝拉拔和微观组织试验。试验结果表明:铸杆微观组织由共晶树枝晶和共晶群体组成。拉拔变形量达98.06%时,枝晶组织基本消失,以变形态的细小等轴晶组织为主,且均匀致密。抗拉强度随线径减小持续增加,当线径小于1 mm时,合金的抗拉强度迅速增大,表现出明显的尺寸效应。确定神经网络模型最优拓扑结构为2-10-1,两种神经元网络模型样本数据的相关系数分别为0.794和0.907,粒子群优化-反向传播(PSO-BP)人工神经网络模型测试样本最大相对误差绝对值为3.3%,能准确预测Cu-Ag合金的抗拉强度。更多还原

【Abstract】 A model of back propagation( BP) and particle swarm optimization( PSO) neural network of Cu-Ag alloy based on size effect were established. The neural network models were applied to Cu-Ag alloy microwire tensile strength prediction. Experimental study of micro-wire drawing of the alloy with different diameter and silver content shows that the cast rod mainly consists of eutectic dendritic microstructure. The dendritic structure disappears when the strain reaches 98.06%. The microstructure of the alloy is dominated by small and homogeneous equiaxed grains. The tensile strength of the alloy increases with decreasing wire diameter.Significant size effect is found. When the wire diameter is smaller than 1 mm the tensile strength of the alloy increases dramatically. The optimal topology of neural network model is 2-10-1. The correlation coefficients of the two neuron network model sample data are 0. 794 and 0. 907 respectively. The maximum relative error absolute value of the PSO-BP artificial neural network model test sample is 3.3% which can predict the tensile strength of the Cu-Ag alloy.更多还原