基于遗传神经网络的多元渣系活度预测模型  

GA-NN-Based Predicting Model of Activity of Multiple Slag System

作　　者：吴令[1] 姜周华[1] 龚伟[1] 李阳[1] 

机构地区：[1]东北大学材料与冶金学院,辽宁沈阳110004

出　　处：《东北大学学报（自然科学版）》2008年第12期1725-1728,共4页Journal of Northeastern University(Natural Science)

基　　金：国家自然科学基金和上海宝山钢铁集团公司联合资助项目(50174012)

摘　　要：基于遗传神经优化BP神经网络权值和阈值建立了多元熔渣活度模型.人工神经网络能实现任意函数逼近,结构简单;遗传算法是建立于遗传学和自然选择原理基础上的一种全局优化搜索算法,能根据个体的适应度函数,通过对个体施加遗传操作实现群体内个体结构重组的迭代处理,逐代演化出越来越好的近似解.通过对CaO-SiO2,CaO-SiO2-Al2O3,CaO-SiO2-Al2O3-MgO渣系组元活度的计算和仿真表明,遗传神经网络具有很强的非线性拟合能力,计算结果在不同的情况下均能很好地吻合文献值,因此能够准确预报多元渣系中组元活度值.A model based on GA-NN for predicting the activity of components in multiple salg system is developed. The artificial neural nets（NN） can implement any approximation of function with simple structure, while the genetic algorith（GA） is a globally optimized search one based on genetics and natural selection theory, which is available to implement the iteration process through allpying the genetic manipulation to the individuals in colonies for their restructuring and then evolve the increasingly improved approximate solutions generation by generation in accordance to the adaptibility function for individuals. GA is always used to give the weights and threholds of neural nets. Computing and simulating the CaO-SiO2,CaO-SiO2-Al2O3,CaO-SiO2-Al2O3-MgO slag systems, it is found that GA-NN model has high nonlinear capability and the computation results fit well with that in relevant earlier works, thus enabling the accurate prediction of the activity of components in molten slag.

关 键 词：熔渣 活度 遗传算法 神经网络 

分 类 号：TF534.1[冶金工程—钢铁冶金] O242.1[理学—计算数学]