【作者】 孙传余；

【导师】 肖林京；

【作者基本信息】 山东科技大学 ， 机械电子工程， 2010， 博士

【摘要】 磁悬浮轴承是利用非接触的磁力作用,将转子悬浮起来,实现定子与转子之间的无摩擦运行。磁悬浮轴承具有无摩擦、噪音小、高转速、寿命长等优点,可应用于风力发电、真空超净环境、高速机床、机械加工、旋转电机、运输医疗等领域。为解决目前磁轴承结构中存在的重力干扰、功耗大、磁路耦合、控制复杂、成本高等问题,本文提出了新型低功耗永磁偏置混合磁轴承的结构。采用永磁“上吸下斥”作用力抵消转子的重力；采用电磁与永磁之间的排斥力调节转子的径向位移,可以按“同增同减法”或PID(Proportion Integration Differentiation)算法调节±X和±Y方向上线圈的控制电流；采用电磁永磁与吸力盘之间的吸引力调节转子的轴向位移,轴向偏置电流为零,能降低电流消耗。在所用材料研究的基础上,推导了与磁轴承结构相符的重力抵消力、径向调节力、轴向调节力表达式,得到了磁悬浮转子单自由度和五自由度数学模型、位移刚度系数、电流刚度系数和S域传递函数,仿真了磁轴承结构中的磁力线分布和五自由度磁悬浮转子运动情况。研究了磁轴承的双闭环控制系统,外环采用无传感器检测的“高频注入参数估计法”测量转子位移,节省电涡流位移传感器,降低成本,内环采用霍尔电流器件ACS712测量线圈电流,提高电流的响应速度。设计了10路电流和5路位移输入电路及10路PWM(Pulse Width Modulation)输出电路,求解了电路的输入输出关系,得到了控制电路的传递函数,得到了双闭环控制系统的传递函数,仿真了传统PID控制下的响应曲线。为适应建模不准确,非线性、不确定、不确知系统的控制需要,研究基于RBF(Radial Basis Function)预测模型的神经网络PID算法,通过神经网络的离线和在线训练,不断修改权值,提高了系统的鲁棒性和自适应性,仿真表明神经网络PID下的系统响应具有更小的超调量和更短的稳定时间。搭建了磁悬浮实验台,绘制了软件流程图,研发了CCS与VC++实验程序代码,得到了实验数据和曲线,实验表明：转子位移波动在±0.1mm,线圈电流波动在±0.4A,实现了对转子的无接触悬浮,且“同增同减”控制下的位移波动更,但电流消耗增大。下一步将继续优化控制策略,减少转子波动和磁轴承功耗,提高整体性能。更多还原

【Abstract】 Magnetic bearing suspendes rotor with noncontact magnetic force, realizes the zerofriction running between rotor and stator. Magnetic bearing has the advantages of zerofriction, low noise, ultrahigh speed and long service life etc, can be applied to the domains of wind power generator, ultraclean vacuum environment, high speed machine, mechanical processing, rotating motors, transport and medicine etc.In order to solve the present problems of magnetic bearing structure, such as gravity disturbance, high power consumption, magnetic path coupling, complicated control, and high cost, etc., the paper puts forward a new structure of low-power hybride magnetic bearing with bias permanent magnet. It adopts the "up-suction and down-repulsion" permanent magnet force to counteract the gravity of rotor; it adopts the repulsion force between electromagnet and permanent magnet to adjust radial displacement of rotor, the coil current of±X and±Y directions can be changed by the method of simultaneous increase or decrease, or by the arithmetic of PID (Proportion Integration Differentiation); it also adopts the suction force among iron plate, electromagnet and permanent magnet to adjust axial displacement of rotor, the bias current is zero at axial direction, thus the current consumption can be lower. On the base of material research, the paper deduces the expression formula of gravity counteract force, radial adjusting force, axial adjusting force, which is consistent with the magnetic bearing structure, then gets the mathematical model, the displacement rigidity coefficient, the current rigidity coefficient, the transfer function of S domain under single degree-of-freedom (dof) and five dof, simulates the flow direction of magnetic force lines in magnetic bearing structure, and simulates the five dof movement of magnetic suspension rotor.The paper also researches the double closed-loop control system of hybrid magnetic bearing, which adopts the no sensor detecting method of "high frequency injection and parameters estimation" to measure the displacement of rotor at outer loop, also adopts the HALL current device ACS712 to measure the coil current at inner loop, thus abandons the displacement sensors of electricity eddy, cuts the cost lower, improves the response speed of coil current. Designes the control circuits of 10 current inputs,5 displacement inputs, as well as 10 PWM (Pulse Width Modulation) outputs, gets the circuit relationship between inputs and outputs, gets the transfer function of control circuit, gets the transfer function of double closed-loop control system, simulates the response curve of traditional PID control system. For the control demand of imprecise mathematical model, non-linear, uncertainty, unknown system, the paper studies on the neural network PID algorithm which is based on RBF (Radial Basis Function) prediction model, through off-line and on-line training of neural network, the weights values are revised continuously, so improves the robustness and adaptability of system, simulation shows that the response curve of neural network PID control system has less overshoot and shorter stable time.The paper builds a magnetic suspension experiment platform, draws the software flowchart, developes the procedure code of CCS and VC++, gets the experimental data and curves, the experiment shows that displacement fluctuations of magnetic suspension rotor is within±0.1 mm, coil current fluctuations is within±0.4A, the raotor has been suspended noncontactly, and the displacement fluctuations is smaller under the control method of simultaneous increase or decrease, the current consumption is higher. Optimizing the control strategy, reducing displacement fluctuation of rotor and current consumption of magnetic bearing, improving overall performance would be the next work.更多还原