转子-轴承系统混沌运动的神经网络反馈控制方法

用神经网络技术对刚性Jeffcott转子-轴承系统进行混沌滞延反馈控制研究。研究结果表明，当转子-轴承系统进入混沌状态后，引入时间滞延反馈控制信号，可以消除转子-轴承系统的混沌振动，使嵌入在混沌吸引子中的不稳定周期轨道回到稳定周期轨道上。采用间接误差计算的BP神经网络学习方法和自适应学习率BP算法结合而形成的改进型BP神经网络方法，可以快速搜寻到次优化的滞延反馈控制强度，从而即时有效地消除转子-轴承系统的混沌振动。一旦混沌振动回归稳态周期振动，则反馈控制信号自动消失。该方法为控制转子-轴承系统的振动状态提供了理论依据，特别是对工程实际转子系统有实用价值。

Feedback control for chaotic motion of a rotor-bearing system with an intelligent neural network

Here,the control of chaotic vibration using a neural network for a rigid Jeffcott rotor system supported on short journal bearings was presented.The study results demonstrated that a feedback control signal is applied to the rotor-bearing system,an unstable periodic orbit embedded in a chaotic attractor is attracted back to a stable periodic orbit;from learning of the neural network,the time-delay feedback control gain is automatically traced.Numerical simulations showed that with a short time,a chaotic vibration is stabilized effectively and then the feedback signal automatically disappears.This method offered a theoretical basis for control of chaotic vibration of a rotor-bearing system,especially,for rotor systems in engineering practice.