基于正交设计和BP神经网络模型的复鞣填充工艺优化

Optimization of Retanning and Filling Process Based on Orthogonal Design and BP Neural Network Model

神经网络模型能够捕捉复杂系统中变量之间可能存在的非线性关系,常用于数据统计和预测中。使用丙烯酸树脂、栲胶、角蛋白/氨基树脂复合材料、合成鞣剂对蓝湿革进行复鞣填充,并利用正交设计和BP神经网络模型分析不同因素对坯革性能的影响。结果表明,以柔软度和撕裂力为评价指标,当中间层节点数为8和6时,神经网络模型可对复鞣填充进行高效拟合,均方误差MSE分别为0.113和0.279,线性拟合相关系数R^(2)分别为0.979和0.955。与正交设计相比,神经网络模型可更好地分析和预测坯革的性能,偏差<5%。

Neural network models can capture the possible nonlinear relationships between variables in complex systems,and have been often used in data statistics and prediction.Wet blue leather was retanned and filled with acrylic resin,tannin extract,keratin/amino resin composite and synthetic tanning agent,and the effects of different factors on the properties of crust were analyzed by orthogonal design and BP neural network model.The results showed that with softness and tearing force as evaluation indicators,when the number of nodes in the middle layer were 8 and 6,the neural network model can efficiently fit the retanning and filling,and the mean square error MSE were 0.113 and 0.279,respectively,the linear fitting correlation coefficients R^(2) were 0.979 and 0.955.Compared with the orthogonal design,the neural network model can better analyze and predict the performance of the leather,with the deviation was less than 5%.