自组织模糊神经网络的内模控制研究与应用  

Research and Application of Internal Model Control Based on Self-built Fuzzy Neural Network

作　　者：王华秋[1] 

机构地区：[1]重庆理工大学计算机科学与工程学院,重庆400054

出　　处：《计算机仿真》2014年第4期343-346,358,共5页Computer Simulation

基　　金：教育部人文社会科学研究青年基金项目(10YJC870037);重庆市教委科学技术研究项目(KJ100805)

摘　　要：研究氧化铝深度脱硅优化问题,为了保证生产质量,对生产过程进行稳定控制,稳定氧化铝深度脱硅过程的热工制度和降低能耗,采用了模糊神经网络辨识的内模控制方法进行氧化铝深度脱硅工艺过程控制,改变了以往依靠人工经验构造控制规则而进行的半自动控制模式。考虑到模糊神经网络隐含层神经元的数目决定了整个网络的规模和性能,根据模糊隶属度函数的激励强度和衰减程度可以添加或者删除模糊神经网络隐含层神经元,从而优化了模糊神经网络隐含层结构,再用自组织模糊神经网络辨识内模控制系统的正模型和逆模型,改进模型的神经网络结构可根据性能要求动态调整,从而改进了神经网络内模控制技术。实验结果表明,新提出的控制方法比传统方法在鲁棒性和抗扰性方面具有更好的性能,各项指标均优于传统控制方法,实现了氧化铝深度脱硅工艺优化。To stabilize the thermal regulation of alumina deep desilication and reduce energy consumption, this paper adopted fuzzy neural network identification into internal model control process control of alumina deep desilica tion. The semi automatic control mode was changed, which relies on artificial experience to construct control rules in the past. Considering that the neurons number in the hidden layer of fuzzy neural network determines the size and performance of the entire network, this paper used the excitation intensity and attenuation degree of fuzzy function to add or delete the hidden layer neurons of fuzzy neural network. Thus the structure of fuzzy neural network hidden lay er was optimized. Then the forward model and inverse model of internal model control system were identified by the self built fuzzy neural network, whose neural network structure can be dynamically adjusted based on performance requirements. Thereby the performance of internal model control was improved. The experimental results show that the proposed control method has better performance than traditional control methods in aspects of robustness and im munity because each algorithm index is better than traditional control method. Hence optimization of deep desilication process has been achieved.

关 键 词：深度脱硅 自组织模糊神经网络 内模控制 系统辨识 

分 类 号：TP273[自动化与计算机技术—检测技术与自动化装置]