欢迎您！东篱公司
退出
-
申报数据库
-
立项数据库
-
成果数据库
-
项目获奖数据库
中文摘要：
针对目前扫描隧道显微镜超高精度位移的要求，提出了一种基于神经网络的精密工作台控制方案。在分析微位移机构工作原理的基础上，建立了工作台的数学模型。神经网络辨识器在线辨识工作台的机械参数，神经网络自学习PID控制器代替传统PID控制器，实现了样本的在线采集，有效克服了神经网络控制器需要离线训练的缺点。利用BP网络的自学习和自适应能力，实时调整网络加权值，改变控制器的控制系数，增强了系统的实时控制性能。实验结果表明，相对于传统PID控制算法，参考位移量为10Fm时，达到稳态值的时间从3．8S缩短到1．8S，稳态误差从4．29／6减小到1．99／6。
英文摘要：
According to the request of ultra-high precision control for stage of STM, a new control scheme based on neural network is proposed. The mathematical model is set up by analyzing the principle of the actuator. Mechanical parameters are identified on-line by the neural identifier. The traditional PID controller is replaced by a neural network self-learning PID controller. The on-line collection of sample is realized. The disadvantage of controller＇s off-line training is overcome. The weights of BP network and the parameters of neural net PID controller can be adjusted by the functions of self-learning and adaptability. The performance of real control system is enhanced. The results of experiment show that the response time is shortened from 3.8 s to 1.8 s, and the stable error is reduced from 4.2 % to 1.9 % under a displacement of 10 μm.
同期刊论文项目
同项目期刊论文
期刊信息
位置：