基于BP神经网络的敏感性分析模型在天然橡胶耐磨性分析中的应用

Application of sensitivity analysis model based on BP neural network in abrasion property analysis of natural rubber composites

采用主成分分析法将天然橡胶复合材料的8种力学性能降维后作为BP神经网络的输入向量,耐磨性作为输出向量,得到训练优秀的BP神经网络。对输入向量进行敏感性分析,得出各主成分对耐磨性影响大小的顺序。对影响最大的主成分所包含的力学性能再一次进行敏感性分析,得到对天然橡胶耐磨性影响最大的力学性能。结果表明,BP神经网络的预测精度达到93.1%,说明BP神经网络适用于橡胶材料耐磨性的预测。

The dimensionality of 8 mechanical properties of natural rubber composites was reduced by principal component analysis, and then the components were used as the input vectors of BP neural network, while abrasion property as the output vector. After training,the BP neural network was established, and the sensitivity matrixes of the input vectors were calculated in order to search the mechani- cal properties which had most significant effect on abrasion property. The results showed that the prediction accuracy of the BP neural network was 93.1% ,indicating that the BP network was suitable for the abrasion prediction of NR composites. Sensitivity analysis indicated that modulus at 100% was the most important factor to influence the abrasion property of NR composites.