基于主成分-神经网络预测干洗后织物复合体粘合效果

Bonding effect prediction of dry-cleaned composite fabric by principal component analysis and neural network

通过主成分分析,提取了面料和粘合衬性能参数的8个主成分作为新的综合变量。采用BP神经网络技术建立预测干洗后织物复合体粘合效果的3层神经网络模型,运用动量法和学习率自适应调整算法训练模型。通过预测值与试验观测值的比较,表明用主成分神经网络方法预测粘合后织物复合体经干洗后粘合效果具有相当高的准确性,从而在一定程度上证明此方法的可行性。

Eight principal components were obtained from the related parameters of the fabric and adhesive lining through principal component analysis, which were used as new variables. The BP neural network technology was adopted to construct a three-layer neural network model for prediction of the bonding effect of dry-cleaned composite fabric, and the model was trained using the vector and the algorithm for learning to adapt to new situations. The comparison of the predicted values and the experiment test values indicated that the prediction of bonding effect of dry-cleaned composite fabric by neural network is rather accurate and this testified in a certain extent that this method is practical.