基于正交设计与BP神经网络优化制备纳米WC-MgO复合粉末

Optimizing preparation of nanocomposite WC-MgO powders based on orthogonal design and backpropagation neural network

为获得晶粒尺寸较小的纳米复合粉末,运用正交实验设计结合BP神经网络优化球磨工艺参数。以磨球直径、球磨转速和球料比为正交实验设计因子,每个因子各取4个水平,以WC-MgO复合粉末的晶粒尺寸为目标因子,编制3因素4水平正交设计表。结合BP神经网络强大的自学习和函数拟合功能,以正交设计表中3因素为网络输入层,以晶粒尺寸为网络输出层,建立BP神经网络优化模型,并通过该模型进行预测和优选,得到最佳的高能球磨工艺参数。

The ball milling processing parameters were optimized by orthogonal design in a combination with BP neural network to get Nanocomposite powders with smaller crystallite size. The three factors four levels orthogonal design table was established with milling ball diameter, milling speed and ball-to-powder weight ratio as factors and crystallite size as the goal factor. The optimal modified ball milling processing parameters were found via predicting and selecting the BP network optimization model with three factors as inputs and crystallite size as output on the basis of the self-learning and effective fitting function.