摘要
分析了BP网络辨识器和控制器的常规训练方法存在的不足之处及其限制，提出了一种通过BP网格辨识器的灵敏度网络反传对象的输出误差来训练BP网络控制器的新方法，根据BP网络和自适应控制的相容性，提出了一种新的BP网络自适应控制结构，并利用BP网络辨识器的灵敏度网络训练BP网络控制器的思想，来调整BP网络控制器的参数，等离子喷徐过程控制模型的训练研究表明其效果良好。
The drawback and limit of the general methods for training BP network identifier and control are analyzed, a new method for training BP network controller by error back propagation through sensitivity network of BP network identifier is put forward. On basis of resemblance between BP network and adaptive control, a new adaptive control structure using BP network is put forward. The parameters of BP network controller is updated by the idea of training BP network controller with error back propagation through the sensitivity network of BP network identifier. The training results of plasma spraying process control model clearly show its excellent effect.
出处
《控制理论与应用》
EI
CAS
CSCD
北大核心
1997年第5期660-664,共5页
Control Theory & Applications
基金
广东省自然科学基金
