摘要
铝液液位和温度、雾化氮气压力和温度以及环境氮气压力和温度之间,具有复杂的非线性、大滞后、强耦合和严重不确定性的特点,很难用机理建模方式获得它们的确定模型。针对这一问题,以优化铝粉粒度分布为生产目标,提出了基于RBF神经网络建立铝粉氮气雾化生产过程模型,采用改进的遗传算法对雾化过程各控制量的设定值进行优化,根据工况条件的变化动态地设定各个控制环节的设定值使铝粉氮气雾化效果达到最佳,从而提高微细铝粉收率的优化控制方法。优化前和优化后的铝粉粒度分布对比结果表明,采用该方法能显著地改善氮气雾化效果,有效提高微细铝粉收率。
The nonlinearities, large time delay, strong coupling and severe uncertainty exists among the melted aluminum level and temperature, atomizing Nitrogen temperature and pressure, and recycle Nitrogen temperature and pressure. It is difficult to obtain the deterministic model by mechanistic method. To the problem, the aluminum powder Nitrogen atomization process model based on RBF neural networks is presented, and enhanced GA based aluminum powder Nitrogen atomizing process optimial control is implemented to improve the percentage of super-tiny aluminum powder. Comparisons of the aluminum powder granularity distribution before and after optimizing show that the optimal control of aluminum powder Nitrogen atomizing process can improve the effect of Nitrogen atomization and promote the percentage of super-tiny aluminum powder greatly.
出处
《控制工程》
CSCD
2007年第5期497-500,504,共5页
Control Engineering of China
基金
国家"十五"科技攻关计划资助项目(2001BA204B01)
高等学校骨干教师计划资助项目(69825106)
关键词
铝粉
氮气雾化
RBF神经网络
过程建模
优化控制
遗传算法
aluminum powder
Nitrogen atomization
RBF neural networks
process modeling
optimal control
genetic algorithms
