摘要
为解决MOCVD设备温度控制的非线性、时变性以及大时滞等问题,提出改进的基于误差反向传播算法(BP)的神经网络控制方法.在传统BP算法基础上附加一个使搜索快速收敛全局极小的惯性项,在微调权值修正量的同时也使学习避免陷入局部最小.该方法不仅具有自学习自适应能力,而且具有自调整比例因子功能.仿真和试验表明,神经网络控制器具有很强的鲁棒性、自学习功能和自适应解耦.在整个温度控制范围基本误差可达到1℃‰,有效的改善MOCVD系统温度的控制性能,对实际温度控制具有较好的指导意义.
A novel PID control based on neural network is presented to deal with the control problem for a nonlinear process with time-delay. The principle of this controller is discussed and learning method based on back - propagation - algorithm is given. This is a novel on -line BP learning algorithm, which uses a conjugate gradient factor to determine the learning direction and uses a bonus - penalty confident to adjust the learning step. This method has not only the learning ability and the adapt ability, but also the self-adjusting factor function, thus allowing for on - line adaptive control strategies to become feasible. Digital simulation and experiment results show that this new controller can improve the dynamic performance of temperature regulation system and robustness. Both simulation and practical application results show that this controller has strong robustness and the capabilities of self -learning and adaptive decoupling. In addition, the basic error can reach to 1℃% referred to input range.
出处
《重庆大学学报（自然科学版）》
EI
CAS
CSCD
北大核心
2006年第8期77-81,共5页
Journal of Chongqing University
基金
陕西省自然基金资助项目(2004F29)
陕西省科学技术研究发展计划项目(2004k05-G1)
关键词
神经网络
BP算法
金属有机化合物化学气相淀积
温度控制
时滞
neural network
the back propagation algorithm （MOCVD）
temperature control
time delay （BP）
Metal Organic Chemical Vapor Deposition
