【作者】 曹淑瑛；

【导师】 王博文；

【作者基本信息】 河北工业大学 ， 电工理论与新技术， 2004， 博士

【摘要】 超磁致伸缩致动器具有应变大、推进力大、快速响应、纳米分辨率等优点，在超精密定位、机器人、减震控制等领域有着广阔的应用前景。然而，超磁致伸缩致动器的外加磁场与输出位移及力存在着显著的磁滞非线性现象，这给其应用带来很大困难。本文选择了“超磁致伸缩致动器的磁滞非线性动态模型与控制技术”这一既具有科学价值又具有工程实际意义的课题，以期实现致动器的控制及使用。 本论文首先对超磁致伸缩材料及其应用现状作了介绍，集中对致动器的模型和控制技术作了全面深入的分析，对神经网络控制的现状和应用前景进行了概述。 随后，基于Jiles-Atherton磁化强度模型、二次畴转模型、非线性压磁方程和致动器结构动力学原理，建立了致动器的磁滞非线性动态模型。应用该模型对致动器的输出应变和力进行计算，计算结果与实验结果符合较好，验证了模型的正确性和实用性。 提出了混合编码遗传算法和信赖域算法相结合的两种混合遗传算法：HGA1和HGA2。就混合遗传算法的关键技术和实现过程做了详细的研究，并将其应用于致动器磁滞动态模型的参数辨识中。仿真和实验研究表明，HGA2性能明显优于HGA1，其能以较快的速度、非常大的概率求得辨识值，并具有一定的抗噪能力。 为了达到纳米分辨率，提出了基于TMS320C31 DSP的致动器位移闭环控制系统。对控制系统的硬件进行了研究，包括TMS320C31控制板，数控恒流源和位移数据采集通道。设计并编写了控制系统软件，实现了致动器系统的自动测试。 设计了一种单参数模糊自整定PID控制系统，并对该控制系统进行了仿真研究。在此基础上，应用单参数模糊自整定法对致动器系统的PID控制器参数进行了整定，编写了PID控制算法软件，实现了系统的微位移自动控制。数字仿真和实验结果证实了该单参数模糊自整定法的有效性和实用性。同时，实验结果和理论分析表明基于TMS320C31 DSP的致动器闭环控制系统具有稳定性好、抗干扰性高的特点，在40μm的量程范围内能达到40nm的分辨率。 由于内在的磁滞非线性，超磁致伸缩致动器总会在开环系统中引起定位误差，在复杂的跟踪问题中造成闭环系统不稳定。为了克服这个问题，根据致动器的磁滞非线性特性，设计了一个动态递归神经网络DRNN，并构造了一种“DRNN前馈+PID反馈”控制方案。仿真研究表明这种控制方案可以在很短的时问内，在线学习建立起致动器磁滞非线性逆模型，消除致动器磁滞非线性的影响，使系统输出较好地跟踪参考输入。更多还原

【Abstract】 Giant magnetostrictive actuators, characterized by large strain, high force, fast response and nanometer solution and so on, have a wide range of potential applications in super-precision positioning, robotics, and vibration control, etc. However, the relation between the input magnetic field and output displacement for magnetostrictive actuators exhibits the hysteresis nonlinearity, which presents a challenge in the applications of actuators. To control and use the actuators, the dynamic model with hysteresis nonlinearity and control technique for giant magnetostrictive actuators are selected as the subject of dissertation for Ph.D.Firstly, giant magnetostrictive materials and theirs applications are introduced, both the models and control techniques for giant magnetostrictive actuators are analyzed in depth, and the actuality and the perspective of applications for neural networks are summarized.Subsequently, the dynamic model with hysteresis nonlinearity for magnetostrictive actuators is founded according to the Jiles-Atherton magnetization model, quadratic moment domain rotation model, nonlinear piezomagnetic equation and actuator structural dynamics principle. The output strain and force for a magnetostrictive actuator has been calculated. It is found that the calculated results are in a good agreement with the experimental ones. This indicates the model’s validity and practicability.Two kinds of hybrid genetic algorithms (HGAs), which are respectively called as the HGA1 and the HGA2, are proposed by combining trust-region algorithm (TRA) with a hybrid coded genetic algorithm. The key techniques and realization processes of the HGAs are thoroughly studied. The HGAs are then applied to identify parameters of dynamic model with hysteresis nonlinearity for magnetostrictive actuators. The simulation and experimental results show that the HGA2 obviously excels the HGA1. The HGA2 can obtain accurate parameters value with a rather convergence speed and a rather large probability, and has the definite ability to resist noise.In order to obtain nanometer resolution, a displacement closed loop control system for the actuator is presented based on TMS320C31 DSP. The control system hardware is studied, including the TMS320C31-based digital signal processor board, constant current source, displacement data collected channel. The control system software is designed and compiled, and the automatic measurement for the actuator system is realized.A kind of single parameter fuzzy self-tuning PID control system is designed, and the simulation study of the control system is carried out. On the base, the single parameter fuzzy self-tuning method is used to tune the PID parameters of the actuator system, the PID control algorithm is compiled, and the displacement automatic control of the actuator system is realized. The numerical simulation and experimental results show the fuzzy self-tuning method’s efficiency and utility. At the same time, the experimental results and theory analysis indicate that the actuator closed-loop control system has good stability and high anti-noise ability, and can achieve 40nm resolution with the range of 40um.Due to the inherent hysteresis nonlinearity, the magnetostrictive actuator always causes position error in the open-loop system, and causes instability of the closed-loop system in complicated tracking problems. In order to remedy this problem, a dynamic recurrent neural network (DRNN) is designed, and a control strategy combining the DRNN feedforward and PID feedback controllers is applied. Numerical simulation results show the control strategy can on-line obtain inverse model of hysteresis nonlinearity for the actuator in very short time and eliminate the impact of hysteresis nonlinearity. Thus, the system output can better track reference input.更多还原