混沌粒子群算法的烧结碳耗BP神经网络模型  

BP neural network model of coke consumption of sintering process based on chaotic PSO algorithm

作　　者：陈鑫[1,2] 翁卫卫 吴敏[1,2] 曹卫华[1,2] 

机构地区：[1]中南大学信息科学与工程学院,湖南长沙410083 [2]先进控制与智能自动化湖南省工程实验室,湖南长沙410083 [3]中南大学信息科学与工程学,院湖南长沙410083

出　　处：《计算机与应用化学》2013年第10期1223-1226,共4页Computers and Applied Chemistry

基　　金：国家自然科学基金重大国际合作研究项目(61210011);国家863计划研究项目(2012AA040307)

摘　　要：有效地计算和预测烧结碳耗,是有针对性地优化烧结生产以降低烧结碳耗的关键前提。文章首先提出了烧结过程碳耗指标——综合焦比并给出合理可行的烧结过程碳耗指标计算模型;再次,结合机理进行分析和灰色关联度分析方法,确定了影响碳耗的主要因素;最后,建立了基于混沌局部搜索粒子群算法的烧结碳耗BP神经网络预测模型(CPSO-BPNN),用带混沌局部搜索的粒子群算法对烧结碳耗BP神经网络预测模型的初始网络权值、阈值进行寻优,以克服BP算法参数寻优时陷入局部极小的缺点。仿真结果表明,CPSO-BPNN可有效地对烧结碳耗进行预测,为优化烧结生产过程,降低烧结碳耗奠定了基础。Efficient calculation and prediction of the coke consumption of the sintering process are pivotal to optimize the sintering process to reduce the coke consumption. A reasonable coke consumption indicator is defined and its calculation mode is developed. Then, the factors affecting the consumption of coke are determined through the integrated use of the mechanism analysis and gray correlation analysis method. At last, a CPSO-BPNN predictive model is developed, in which the particle swarm optimization combined with chaotic local search is used to optimize the initial connection weights and translate scaling factors of the BP neural network model. The simulation result demonstrates that CPSO-BPNN provides an effective way to predict the coke consumption, which serves as a basis to optimize sintering process and reduce the coke consumption.

关 键 词：烧结碳耗 混沌局部搜索 粒子群优化 BP神经网络 

分 类 号：TF512[冶金工程—钢铁冶金]