BP神经网络在油脂亚临界水解反应中的应用

亚临界水解反应制备脂肪酸的研究方法大多以试验为主,得到一些规律性的结论。采用BP神经网络预测橡胶籽油水解反应的水解率。结合均匀试验方法得到神经网络所需的训练样本,确定最佳的网络结构为3-8-1。并检验了所得网络结构的泛化能力,通过线性回归分析得到网络的输出与试验数据的相关系数达到了0.994,证明所确定的网络结构在油脂亚临界水解反应中的预测效果较好。

Application of BP neural network in subcritical hydrolysis of oil

Up to the present,the subcritical hydrolysis of oils to fatty acid is studied mainly by experiments,and some regular conclusions have been obtained from these experiments.In this paper,BP neural network was used to forecast the hydrolysis rate of rubber seed oil subcritical hydrolysis reaction.The training samples of BP neural network were obtained by uniform design experiments,and the optimized network structure was determined as 3-8-1.The generalization ability of network was examined,and the linear regression analysis showed that the correlation coefficient between the network output and the experimental data reached 0.994,which indicated that the determined network structure could obtain good results in forecasting the hydrolysis reaction.