基于BP神经网络的色纺纱配色

针对传统配色模型实用性差的问题，利用神经网络强大的非线性映射能力，探讨基于人工神经网络的色纺纱配色方法。构建了色纺纱BP神经网络配色模型，研究了多种BP算法，隐含层节点数对仿真效果及泛化能力的影响。实验结果表明:基于BP 神经网络的色纺纱配色方法可以实现色纺纱反射率与配方之间的非线性映射,新型算法（Levenberg-Marquardt、拟牛顿、共轭梯度算法）在迭代次数和仿真时间上有较大的优势，隐含层节点数对仿真结果影响较小，训练样本的平均预测色差DEcmc(2:1)为0.18，但训练样本范围外的预测色差较大，因此提高该神经网络的泛化性能是下一步研究的关键。

Recipe prediction of melange yarn based on BP neural network

In view of poor practicability of conventinal color matching models of melange yarn, a new method of color matching based on the powerful nonlinear mapping capability of neural network was discussed. In the article, color matching model of melange yarn based on BP neural network was built, the effects of BP algorithms and nodes number of hidden layer on simulative effect and generalization were studied. The results show that, the color matching model of melange yarn based on BP neural network can achieve the nonlinear mapping between the reflectivity of melange yarn and the recipes. The new algorithms such as Levenberg-Marquardt, Quasi-Newton and Conjugate Gradient Algorithms have greater advantages on the iterations and simulation time. And the nodes number of hidden layer has little impact on the result of the simulation. The average predicted color difference DEcmc(2:1) of the training sample is 0.18, but outside the scope of the training sample is bigger, which means the generalization of this color matching model based on BP neural network need to enhance in the further study.