论文部分内容阅读
神经网络偏最小二乘法 (NNPLS)被应用于一种甲烷氧化偶联多组分催化剂的鲁棒反应模型的建立 .重点研究了内层神经网络学习算法、激活函数、网络结构 (包括隐含节点数、隐含层 )、网络权值初始化及主元的选取原则等 .研究表明 ,内层神经网络分别采用 1 10 5 1,1 8 4 1,1 8 5 1,1 7 4 1,1 8 4 1,1 8 6 1的拓扑结构是合适的 ;Levenberg Marquardt方法被用于网络的学习算法可以加快学习速度 ;同时采用了sigmoid函数为激活函数 .计算结果显示 ,四主元可以满足建模的需要 .与单纯的神经网络催化剂模型相比 ,NNPLS方法压缩分解了变量 ,减少了计算量 ,同时使模型的推广能力得到提高 ,有效地改善了直接神经网络建模过程中催化剂模型泛化能力较差的缺点 .
Neural Network Partial Least Squares (NNPLS) has been applied to a robust reaction model of methane oxidative coupling multicomponent catalysts. The study focused on the inner neural network learning algorithm, activation function, network structure (including hidden nodes Number, hidden layer), initialisation of network weights and principle of principal component selection, etc. The research shows that inner neural networks use 1 105, 1 184, 1 85 1, 1 7 4 1, 8 1 4 1,1 8 6 1 topology is appropriate; Levenberg Marquardt method is used in the network learning algorithm can speed up the learning speed; At the same time, the sigmoid function is used as the activation function. The calculated results show that the four principal elements can satisfy the modeling Need.Compared with pure neural network model, NNPLS method compresses and decomposes variables, reduces the computational complexity and improves the generalization ability of the model, which effectively improves the generalization ability of the catalyst model in direct neural network modeling Poor shortcomings.