基于支持向量机的磁力轴承控制算法研究

发布时间：2018-06-28 19:24

本文选题：磁力轴承 + 控制系统　；参考：《武汉理工大学》2011年硕士论文

【摘要】：主动磁力轴承作为一种优秀的机电综合体,它具有许多老式的接触式轴承所不具备的优点,比如没有摩擦,故没有磨损,无需在轴承转子和定子之间涂润滑剂,因此转子运动更快,使用寿命更长。正因为这些优点,主动磁力轴承受到工业领域比如轴承行业以及学术领域的广泛关注和热议。但由于磁力轴承本身固有的特性,如不稳定性、参数不确定性、模型存在非线性等。在过往的研究中发现,采用传统的PID控制器无法达到理想的控制要求。需将新的算法加入其中进行分析研究。本文主要针对磁力轴承中的单自由度磁力轴承进行讨论分析。在对磁力轴承电磁力和受力问题的分析后,对磁力轴承非线性特性进行建模。然后在传统PID闭环控制的基础上,加入BP神经网络算法和支持向量机算法对PID的控制参数进行调整,通过仿真实验对比两种算法的控制效果。文章介绍了神经网络的学习规则,利用神经网络的高度非线性映射能力,分析设计了BP神经网络PID控制器,仿真结果表明,BP整定PID控制可以有效地减小超调,增加转子的起浮位置。但因为神经网络存在局部极小、易出现过拟合等问题,随着隐含层节点数目的增加,控制性能反而变差。为避免神经网络的缺点,文章提出了基于支持向量机的磁力轴承PID控制。在对支持向量机的基本理论及其回归算法进行了详细介绍后,首先利用支持向量机能逼近任意非线性函数的特点,在传统PID闭环控制的前提下,对磁力轴承的非线性系统进行辨识。然后推导出基于支持向量机的PID控制器算法,结合辨识模型,利用Simulink中的M函数和SVM工具箱实现基于支持向量机PID控制的磁力轴承控制系统的仿真实验。将其与BP神经网络整定PID控制和传统PID控制相比较,仿真结果表明,基于支持向量机的自适应PID控制器的控制效果更好,不仅可以使磁力轴承在更宽范围内起浮,而且调节时间快。
[Abstract]:As an excellent electromechanical complex, active magnetic bearings (AMB) have many advantages that old contact bearings do not have, such as no friction, no wear, no need to apply lubricant between the rotor and stator of the bearing. As a result, the rotor moves faster and has a longer service life. Because of these advantages, active magnetic bearings (AMB) have attracted wide attention and heated discussion in industry such as bearing industry and academic field. However, due to the inherent characteristics of magnetic bearings, such as instability, parameter uncertainty, nonlinear model and so on. In the past research, it is found that the traditional pid controller can not meet the ideal control requirements. The new algorithm should be added to it for analysis and research. In this paper, the single degree of freedom magnetic bearings in magnetic bearings are discussed and analyzed. After analyzing the electromagnetic force and force problem of magnetic bearing, the nonlinear characteristics of magnetic bearing are modeled. Then, on the basis of traditional pid closed-loop control, BP neural network algorithm and support vector machine algorithm are added to adjust the pid control parameters, and the control effects of the two algorithms are compared by simulation experiments. In this paper, the learning rules of neural network are introduced. The BP neural network pid controller is analyzed and designed by using the high nonlinear mapping ability of neural network. The simulation results show that BP tuning pid control can effectively reduce overshoot. Increase the float position of the rotor. However, due to the problem of local minimization and over-fitting of neural networks, the control performance becomes worse with the increase of the number of hidden layer nodes. In order to avoid the shortcoming of neural network, a magnetic bearing pid control based on support vector machine (SVM) is proposed in this paper. After introducing the basic theory of support vector machine and its regression algorithm in detail, firstly, using the feature of support vector function to approximate any nonlinear function, under the premise of traditional pid closed-loop control, The nonlinear system of magnetic bearing is identified. Then, the pid controller algorithm based on support vector machine is deduced, and the simulation experiment of magnetic bearing control system based on support vector machine pid control is realized by using M function and SVM toolbox in Simulink combined with identification model. Compared with BP neural network tuning pid control and traditional pid control, the simulation results show that the adaptive pid controller based on support vector machine has better control effect, and can not only make the magnetic bearing float in a wider range. And the adjustment time is fast.
【学位授予单位】：武汉理工大学
【学位级别】：硕士
【学位授予年份】：2011
【分类号】：TH133.3

【引证文献】