基于遗传算法与神经网络的选厂预测建模研究  

A Research on Modeling Mineral Separation Plant Based on Genetic Algorithms and Artificial Neural Network

作　　者：吴大宏[1] 李英龙[1] 

机构地区：[1]昆明理工大学资源开发工程系,云南昆明650093

出　　处：《中国锰业》2000年第1期24-27,共4页China Manganese Industry

基　　金：云南省中青年学术和技术带头人培养经费资助!项目 (编号 :97E0 4 9Q)

摘　　要：建立了一个由遗传算法和BP网络组成的组合模型GA -ANN。该模型先用遗传算法优化神经网络权空间 ,若网络达不到精度要求 ,再在此基础上 ,用BP算法进一步对权空间进行优化。将该组合模型用于对某选矿厂的实测数据预测建模的初步研究 ,结果表明 :其预测效果好于传统的BP网络和采用“早终止”技术的改进BP网络。A hybrid model named GA-ANN was set up, which adopted Genetic Algorithms and the backpropagation learning of Artificial Neural Network. In the model, Genetic Algorithms was used in optimizing the weights of the multilayer feedforward (MLFF) network first. If the MLFF network can′t meet the requirement of precision, the backpropagation learning was used in furthe optimizing the weights of the MLFF network obtained from the above step. GA-ANN has been used in making an initial research on modeling a mineral separation plant. The results indicated that its prediction precision is better than those of the MLFF network with the conventional backpropagation learning or with the hybrid learning method of the backpropagation learning and the ‘early stopping’ method.

关 键 词：选矿厂 遗传算法 人工神经网络 预测建模 

分 类 号：TD928.9[矿业工程—选矿] TP183[自动化与计算机技术—控制理论与控制工程]