基于BP神经网络的磁悬浮飞轮控制  

Control of Magnetic Suspended Flywheel Using BP Neural Network

作　　者：陈小飞[1] 吉莉[1] 刘昆[1] 

机构地区：[1]国防科技大学航天与材料工程学院,长沙410073

出　　处：《航天控制》2010年第5期3-8,共6页Aerospace Control

摘　　要：针对磁悬浮飞轮控制对鲁棒性、低功耗及不平衡振动抑制等要求,提出BP神经网络直接控制方法。设计了两层结构BP神经网络控制器,基于磁轴承电磁力方程推导网络权值更新算法,实现了神经网络的在线训练。仿真表明,控制器权值更新算法对环境变化适应能力强,训练成功率高;BP神经网络控制器具有起浮迅速、抗干扰能力强、功耗低等性能,并具备不平衡振动抑制能力。结果表明BP神经网络控制器满足磁悬浮飞轮控制要求,具有可行性和有效性。A BP neural network（NN） controller is proposed to meet the control requirements of magnetic suspended flywheel（MSF）,such as robustness,low power consumption and unbalance compensation.A two-layer BP NN controller is designed,from which the weights updating is derived based on magnetic force equation,and the training of NN controller is performed online in the control loop.The simulation shows that the weight updating algorithm is robust against disturbance and guarantees the training ratio,and the rapid response and robustness is achieved by the proposed controller.Moreover,unbalance vibration is eliminated under the constraints of power consumption.The results show that the BP NN controller meets the control requirements of MSF and has feasibility and effectiveness.

关 键 词：磁悬浮飞轮系统 磁轴承 BP神经网络 不平衡振动 

分 类 号：TP229[自动化与计算机技术—检测技术与自动化装置] V448[自动化与计算机技术—控制科学与工程]