基于RBF神经网络模型和优劣解距离分析的复鞣填充工艺优化

Optimization of Retanning and Filling Process Based on RBF Neural Network Model and TOPSIS Method

使用正交设计、RBF神经网络模型、皮尔逊相关性分析和优劣解距离分析等4种统计学方法对皮革复鞣填充进行设计、拟合、分析与优化,并得出以下结论:RBF神经网络模型可很好地拟合制革复鞣填充过程,物理力学性能的均方误差<0.05;皮尔逊相关性分析则可揭示不同皮革化学品与物理力学性能的相关性和显著性,其中羊毛脂与物理力学性能呈现正相关的强显著性(p<0.05);结合RBF神经网络和优劣解距离分析可对全面实验进行预测和评价,从而得到最优解,理论最优解和实测值偏差<10%。

Four statistical methods including orthogonal design, RBF neural network model, Pearson correlation analysis and TOPSIS analysis were used to design, fit, analyze and optimize leather retanning and filling process, and the following conclusions were drawn: RBF neural network model can well fit the leather retanning and filling process, and the mean square error of physical mechanical properties was less than 0.05;Pearson correlation analysis can reveal the correlation and significance of different leather chemicals and physical mechanical properties. The physical mechanical properties showed a strong significant positive correlation with Lanolin(p<0.05);combined with the RBF neural network and TOPSIS analysis, the comprehensive experimental prediction and evaluation could be used to obtain the optimal solution, and the deviation between the theoretical and the measured value was less than 10%.