改进Back Propagation神经网络预测麻纤维/UP复合材料的界面性能
Prediction of bast fiber/UP composites interfacial property by improved Back Propagation neural network

为研究麻纤维化学成分对其增强复合材料界面性能的影响, 选取麻纤维纤维素、半纤维素、果胶、木质素、水溶物、脂蜡质成分含量及回潮率作为影响因素, 以麻纤维/不饱和聚酯树脂(UP)复合材料界面性能作为影响结果, 构建Back Propagation (BP)神经网络的训练样本.首先, 利用灰关联分析法对影响麻纤维/UP复合材料界面性能的因素进行关联度计算; 其次, 按照影响程度的大小进行排序, 建立3层BP神经网络模型进行迭代训练; 最后, 预测麻纤维化学成分含量对麻纤维/UP复合材料界面性能的影响.预测结果表明: 学习结束后模型的输出比较接近实测值, 说明BP神经网络具有很强的学习能力, 同时也证明了将BP神经网络用于麻纤维/UP复合材料界面剪切力预测的可行性;灰关联与BP神经网络联用后预测精度得到大大提高, 预测误差最大可减小83.28%. In order to investigate the influence of fiber chemical composition on interfacial properties of the reinforced composites, the back propagation (BP) neural network training sample was constructed with the composition cellulose, hemicellulose, pectin, lignin, water-soluble material, grease waxand moisture regain of bast fibers as factors, and the interfacial properties of bast fibers/unsaturated polyester resin (UP) composites as the results. The gray relational analysis method was employed to investigate the correlation degree of factors which have influence on the interfacial properties of bast fibers/UP composites, and the factors were ranked according to the size of influence degree.A three layer BP neural network model was constructed for iterative training. The effects of chemical composition content on interfacial properties of bast fibers/UP composites can be predicted. The prediction results show that the prediction model output is close to the measured values after learning. It proves that the BP neural network has strong ability to learn which means that the BP neural network can be used in the prediction of interfacial shear force of bast fibers/UP composites. The prediction accuracy is greatly improved and can be reduced by as large as 83.28% with the combination of gray relational analysis and BP neural network. 国家自然科学基金(51303131,51206122)