【作者】 张森；

【导师】 陈虹；

【作者基本信息】 吉林大学 ， 控制理论与控制工程， 2018， 硕士

【摘要】 为了在地面模拟航天器的空间运动,研究人员提出可以将航天器的空间运动等效变换为地面参考运动,进而可以通过满足给定位置和速度需求的地面移动基座的轨迹跟踪控制,来实现航天器轨道运动的地面模拟,从而降低空间运动模拟的成本,提升开发效率。本文针对移动基座系统底层执行电机的转速跟踪控制展开研究。由于移动基座在运动过程中需要同时满足给定的位置和速度需求,因此,电机在低速区的运动是很难避免的,但是有刷直流电机在低速区的高精度转速控制较难实现,这主要是由于在低速区扰动对电机的转速跟踪产生了不利影响。因此基于移动基座执行器驱动电机宽调速、高精度的需求,提出了基于强化学习的有刷直流电机自适应鲁棒非线性控制方法,以此解决有刷直流电机系统的低速扰动抑制问题,提升转速的跟踪精度。首先,根据有刷直流电机的结构及原理,给出了包含电枢回路模型和刚体动力学模型的电机动力学模型,分析了有刷直流电机低速波动及动态过程中产生死区、爬行等现象的原因。针对LuGre摩擦力矩的非线性特性,为了避免观测器在高转速工况下估计LuGre动态摩擦力模型未知状态时产生的失稳现象,改进了LuGre动态摩擦力模型,使其在高低转速下都能够很好的描述摩擦行为,并且能够避免观测器的失稳现象;针对齿槽转矩的非线性特性,将齿槽转矩用占据主导作用的一阶谐波分量代替,通过快速傅立叶变换确定了齿槽转矩基频,推导具有仿射形式的齿槽转矩模型。进而,建立了面向控制的电机非线性动力学模型。然后,针对有刷直流电机系统的非线性、参数不确定性,提出了基于强化学习的有刷直流电机自适应鲁棒非线性控制方法。为了精确补偿有刷直流电机的齿槽转矩和摩擦力矩,考虑到模型参数的不确定性,基于自适应鲁棒控制方法设计了参数自适应律,在线辨识模型参数。针对系统中存在的建模误差和系统负载等未知扰动,采用基于强化学习Actor-Critic方法的扰动估计方法,在线估计未知扰动。利用有刷直流电机电流环、转速环系统的微分平坦特性,设计了基于微分平坦的二自由度电流环和转速环控制器。最后,为了验证本文所设计有刷直流电机转速控制方法的有效性和优越性,基于dSPACE的电机快速原型试验平台,设计了电机系统的相关硬件电路。通过不同的稳态及瞬态工况下的实验,验证了本文方法的有效性。并且进行了与双环PID控制方法、自适应鲁棒控制方法的控制性能对比。实验结果表明,本文所设计的基于强化学习的有刷直流电机自适应鲁棒控制方法具有更好的稳态及瞬态控制性能。更多还原

【Abstract】 In order to simulate the spacecraft’s spatial motion on the ground simulation,the researchers proposed that the spacecraft’s space motion can be equivalently transformed into the ground reference motion,and can then be tracked and controlled by a ground moving base that satisfies a given position and velocity requirement.To achieve ground simulation of spacecraft orbital motion,thereby reducing the cost of space motion simu-lation,and improve development efficiency.In this paper,the rotational speed tracking control of the bottom motor of the mobile base system is studied.Because the moving base needs to satisfy the given position and speed requirements at the same time during the movement,the movement of the motor in the low speed zone is difficult to avoid,but the high-precision speed control of the brushed DC motor in the low speed zone is difficult to achieve.This is mainly due to the fact that disturbances in the low speed zone adversely affect the motor speed tracking.Therefore,based on the requirements of wide-speed and high-precision drive motors for mobile pedestal actuators,an adaptive robust nonlinear control method for brushed DC motors based on reinforce-ment learning was proposed to solve the problem of low-speed disturbance suppression in brushed DC motors.Improve the tracking accuracy of the speed.Firstly,based on the structure and principle of brushed DC motor,the dynamic model of the motor including armature circuit model and rigid body dynamics model is given.The phenomena of dead zone,crawling and so on in the low-speed fluctuation and dynamic process of brushed DC motor are analyzed.s reason.In view of the nonlinear characteristics of LuGre friction torque,in order to avoid the instability phenomenon of the observer when estimating the unknown state of the LuGre dynamic friction model under high-speed operating conditions,the LuGre dynamic friction model was improved to enable it to operate at both high and low speeds.The friction behavior is well described and observer instability can be avoided;the cogging torque is replaced by the dominant harmonic component of the first order harmonic component of the cogging torque and determined by the Fast Fourier Transform.The cogging torque base frequency deduces an affine cogging torque model.Then,a control-oriented motor nonlinear dynamic model was established.Then,according to the nonlinearity and parameter uncertainty of the brushed DC motor system,an adaptive robust nonlinear control method for brushed DC motor based on reinforcement learning is proposed.In order to accurately compensate the cogging torque and friction torque of the brushed DC motor,taking into account the uncertainties of the model parameters,a parameter adaptive law was designed based on the adaptive robust control method,and the model parameters were identified online.For the unknown disturbances such as modeling error and system load in the system,the disturbance estimation method based on the reinforcement learning Actor-Critic method is adopted to estimate the unknown disturbance on line.Based on the differential flatness characteristics of brushed DC motor current loop and speed loop system,a two-degree-of-freedom current loop and speed loop controller based on differential flatness is designed.Finally,in order to verify the effectiveness and superiority of the design method of brushed DC motor speed control in this paper,the relevant hardware circuit of the motor system was designed based on dSPACE’s motor rapid prototype test platform.Exper-iments in different steady state and transient conditions have verified the effectiveness of the proposed method.And compared with the control performance of dual-loop PID control method and adaptive robust control method.The experimental results show that the adaptive robust control method of brushed DC motor based on reinforcement learning designed in this paper has better steady state and transient control performance.更多还原