【摘要】 在由三种环氧树脂(EP)及两种固化剂组成的胶粘剂体系中,以胶粘剂组成为输入参数、室温剪切强度为输出参数,通过对样本组的训练,得到训练好的神经网络。利用该神经网络可以预测任一给定配方胶粘剂的室温剪切强度,评估单一组分对胶粘剂性能的影响,给出任一性能范围对应的组成分布域,结合遗传算法可以预测最佳配方组成。结果表明:对复杂的胶粘剂体系而言,神经网络和遗传算法可以大大减少试验次数、节省时间且可预测出最佳试验条件等;神经网络和遗传算法在胶粘剂设计中具有广泛的应用前景。更多还原

【Abstract】 In an adhesive system consisting of three kinds of epoxy resin(EP) and two kinds of curing agents,a trained neural network was obtained through the training of sample group by using adhesive composition as the input parameters,and shear strength at room temperature as the output parameters.Using this network,the shear strength at room temperature of a given adhesive formula can be predicted,the influences of a single component on the adhesive properties can be estimated,the corresponding composition distribution area of a certain properties area can be obtained.And the optimal adhesive formula can also be predicted by artificial neural network and genetic algorithm.The results showed that for complex adhesive system neural network and genetic algorithm can reduce experiment number,save time,and predict optimum experiment conditions.Thus neural network and genetic algorithm have broad application prospect in the adhesive design.更多还原