多跨转子系统多频传递力变步长神经网络控制

Variable-step Neural Network Control for Multi-frequency Transmission Forces of Multi-span Rotor Systems

转子在旋转过程中,转子上的各种激励力会通过轴承传递给基础,从而导致基础的振动。为了抑制多频传递力,提出一种基于跟踪滤波后轴颈位移和传递力误差信号的变步长神经网络算法。设计一种集电磁执行器与固定瓦滑动轴承于一体的混合支承结构,并建立其动力学模型。用有限元法建立多跨转子系统的动力学模型,分析传递力控制的基本原理,采用离散跟踪微分器和状态观测器分别解决位移信号夹杂白噪声不易微分和传感器与执行器不同位时执行器位置处轴颈位移的重构问题。提出一种能够抑制多跨转子系统多频传递力的变步长神经网络算法。以一个四轴承二跨转子系统的仿真模型为例进行仿真分析,对所提出的变步长神经网络算法抑制双跨转子系统传递力的有效性进行了验证。结果表明,变步长神经网络算法能够有效地抑制转子系统对基础的多频传递力。

When the rotor system is in operation,the vibration of the base will be caused by the rotor’s multi-frequency excitation forces which transmitted to the base through support bearings.A variable-step neural network algorithm based on tracking filtered journal displacements and transmission forces error signals is proposed to suppress the multi-frequency transmission forces.Firstly,a hybrid bearing integrating electromagnetic actuator and journal bearing is designed and its dynamic model is established.Secondly,the dynamic model of a multi-span rotors system is built by the finite element method,and the theoretical principle of reducing the transmission forces is analyzed.The discrete tracking differentiator and the state observer are used to solve the problem that the displacement signal is not easy to differentiate with white noise and reconstruction of the journal displacement at the actuator position when the sensor and actuator are non-collocated.Then a variable-step neural network algorithm for suppressing the multi-frequency transmission forces from the multi-span rotor system is proposed.Finally,numerical simulations are carried out in a four-bearing double-span rotor system to verify the effectiveness of variable-step neural network algorithm for suppressing the multi-frequency transmission forces from the rotor system to the base.The simulation results show that the variable-step neural network algorithm can effectively suppress the multi-frequency transmission forces from the rotor system to the base.