关键词: 混合滑动轴承, 非圆轴心轨迹, 径向基函数神经网络, MRANEKF算法

Abstract:

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.

Key words: hybrid sliding bearing, non-circular journal locus, radial basis function(RBF) neural network, MRANEKF algorithm