混合轴承-转子运动轨迹的RBF神经网络控制

Controlling Journal Locus in Hybrid Bearing System Using RBF Neural Network

提出了动静压混合滑动轴承支承-转子运动轨迹控制技术和方法,以期用于转子振动控制和非圆截面零件加工。通过调节主动节流器的参数,改变轴承油腔的压力、流量,影响轴承封油面上的压力分布,控制转子按预定轨迹运动。建立了轴承-转子系统的动力学模型,给出了基于最小资源分配(MRAN)和扩展卡尔曼滤波(EKF)的径向基函数(RBF)神经网络自适应控制策略,在线辨识系统参数,以适应其非线性、参数未知特点。计算结果表明,RBF神经网络控制器具有良好的非线性逼近能力;转子不平衡响应或外部同步激励力引起的振动得到了有效抑制,得到了较为精确的转子非圆轨迹跟踪结果。

Based on active hybrid sliding bearing-rotor system,the control technique of the motion of the rotor was studied,which then was used for vibration control and non-circular machining.By adjusting the parameters of active restrictors,the pressure and liquid flow injected into the recesses and the pressure distribution on the lands were altered,and the rotor was forced to move following a reference trajectory.As the dynamic modal of the bearing-rotor system was constructed,an adaptive RBF（radial basis function） neural network called minimal resource allocating network（MRAN） controller combined with extended Kalman filter,that is MRANEKF,was developed to identify the system online and cope with its non-linearity and parameter uncertainty.The simulation results show that the RBF network does well in nonlinear function approximation.The rotor vibration excited by mass imbalance or external synchronous forces can be reduced greatly,and an accurate non-circular reference orbit tracking result can be obtained as well.