基于MATLAB的BP神经网络在旋流器模拟设计中的应用  

Application of BP Neural Network Based on MATLAB Package to the Prediction Performance of Hydrocyclone

作　　者：王志斌[1] 陈文梅[2] 褚良银[2] 王升贵[2] 

机构地区：[1]四川化工职业技术学院,四川泸州646005 [2]四川大学,四川成都610065

出　　处：《流体机械》2007年第10期20-24,共5页Fluid Machinery

基　　金：国家自然科学基金项目(20376049)

摘　　要：利用MATLAB工具箱中的BP神经网络模型建立了旋流器的三层神经网络,提出了旋流器隐层单元数的确定方法,并通过所选择的函数及所构建的网络结构,对网络学习训练得到了14—20—2的预测性能模型,利用所收集的60组样本数据进行训练的结果表明所采用函数和网络结构具有较高精度,由训练得到的权值和阈值对旋流器性能进行预测的结果显示它能满足工程需要。The three layer neural network model of hydrocyclones was created by BP method of MATLAB package, and the method of determining hidden layer cell number was put forward. By choosing calculation function sets and neural network structure, the prediction performance model for 14-20-2 was gotten by training the network from 60 sets collection appropriate teaching data. The result showed that the network could achieve higher precision. Based on the above, the performance of hydrocyclones was simulated by obtaining weight and bias value, the result showed that the created network structure can satisfy engineering demands.

关 键 词：水力旋流器 MATLAB软件包 神经网络 模拟设计 预测性能 

分 类 号：TQ028.5[化学工程] TP183[自动化与计算机技术—控制理论与控制工程]