基于零相位滤波的迭代学习控制器设计
发布时间：2019-05-23 02:30
【摘要】：智能手机行业正方兴未艾。其中,金属手机外壳以其优异的质感、耐磨损等优点越来越受大众的青睐,市场规模巨大。而目前智能手机的金属外壳加工所使用的数控机床为传统机床,并未考虑到同一工件大批量重复加工问题,仍然将大批量的加工任务当作单件来生产,导致次品率居高不下。为解决上述问题,本课题提出了迭代学习加工方法,使机床能够利用先前的加工过程来优化后续加工,以提高加工精度。迭代学习控制为核心控制算法,本课题将对其收敛性、稳定性以及工程应用过程中会出现的过冲现象进行研究,并提出相应解决方案。课题将驱动器调节成速度闭环模式,而在控制器中完成位置环的搭建。基于dSPACE1103控制卡,课题设计了速度环路扫频辨识实验,得到了速度闭环的数学模型。工程实践中往往将位置环的控制模式设为比例(Proportion)控制,为使得被控系统获得良好的动态特性,需要对控制参数P进行适当的优化。课题中将采用迭代反馈整定方法进行控制参数整定。迭代学习控制在应用过程中往往会出现误差先收敛后发散最终又收敛这一过冲现象。课题分别从时域和频域两个角度对这一现象进行分析,得出结论:误差频谱中的高频成分积累最终导致过冲现象。为解决这一问题,需要设计相关滤波器将误差高频成分过滤掉。而普通滤波器存在相位差,于是我们引入零相位滤波器来完成滤波。误差的超前补偿可以使误差更快速地收敛到更小值。无论是零相位滤波还是误差超前补偿都是在两次迭代运行之间的离线状态下完成的。具体的最优超前补偿步数与其对应的最大截止频率,需要依据算法设计相关的仿真实验来确定。根据过冲现象的成因分析和相关的滤波与补偿机制,课题进行了仿真实验,得到了期望效果,误差实现了单调收敛。基于理论推导与仿真实验,课题设计了具体的数控机床实验,误差实现了单调收敛,最终误差可以收敛到1μm以内,验证了零相位滤波与误差超前补偿的效果。并对迭代运行过程中的误差进行了频谱分析,从而可以直观体现出零相位滤波的效果。
[Abstract]:The smartphone industry is in the ascendant. Among them, metal mobile phone shell is more and more popular because of its excellent texture, wear resistance and other advantages, and the market scale is huge. At present, the CNC machine tool used in the metal shell processing of smart phone is a traditional machine tool, which does not take into account the problem of mass repeated machining of the same workpiece, and still produces a large number of processing tasks as a single piece, resulting in a high rate of defective products. In order to solve the above problems, an iterative learning machining method is proposed in this paper, which enables the machine tool to optimize the subsequent machining by using the previous machining process in order to improve the machining accuracy. Iterative learning control is the core control algorithm. In this paper, the convergence, stability and overshoot phenomenon in the process of engineering application will be studied, and the corresponding solutions will be put forward. In this paper, the driver is adjusted to the speed closed loop mode, and the position loop is built in the controller. Based on dSPACE1103 control card, the frequency sweep identification experiment of speed loop is designed, and the mathematical model of speed closed loop is obtained. In engineering practice, the control mode of position loop is often set to proportional (Proportion) control. In order to obtain good dynamic characteristics of the controlled system, it is necessary to optimize the control parameter P. In this paper, the iterative feedback tuning method will be used to adjust the control parameters. In the application process of iterative learning control, the error converges first, then diverges, and then converges. This paper analyzes this phenomenon from the perspective of time domain and frequency domain, and draws the conclusion that the accumulation of high frequency components in the error spectrum eventually leads to overshoot. In order to solve this problem, it is necessary to design a correlation filter to filter out the high frequency components of the error. However, the common filter has phase difference, so we introduce zero-phase filter to complete the filtering. The advance compensation of the error can make the error converge to a smaller value more quickly. Both zero-phase filtering and error advance compensation are completed in the offline state between the two iterations. The optimal number of advance compensation steps and the corresponding maximum cutoff frequency need to be determined according to the simulation experiments designed by the algorithm. According to the cause analysis of overshoot phenomenon and the related filtering and compensation mechanism, the simulation experiment is carried out, and the expected effect is obtained, and the error realizes monotone convergence. Based on the theoretical derivation and simulation experiment, a concrete NC machine tool experiment is designed. The error converges monotonously, and the final error can converge to less than 1 渭 m. The effect of zero phase filtering and error advance compensation is verified. The spectrum analysis of the error in the iterative operation process is carried out, so that the effect of zero-phase filtering can be intuitively reflected.
【学位授予单位】：哈尔滨工业大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP273
本文编号：2483517
[Abstract]:The smartphone industry is in the ascendant. Among them, metal mobile phone shell is more and more popular because of its excellent texture, wear resistance and other advantages, and the market scale is huge. At present, the CNC machine tool used in the metal shell processing of smart phone is a traditional machine tool, which does not take into account the problem of mass repeated machining of the same workpiece, and still produces a large number of processing tasks as a single piece, resulting in a high rate of defective products. In order to solve the above problems, an iterative learning machining method is proposed in this paper, which enables the machine tool to optimize the subsequent machining by using the previous machining process in order to improve the machining accuracy. Iterative learning control is the core control algorithm. In this paper, the convergence, stability and overshoot phenomenon in the process of engineering application will be studied, and the corresponding solutions will be put forward. In this paper, the driver is adjusted to the speed closed loop mode, and the position loop is built in the controller. Based on dSPACE1103 control card, the frequency sweep identification experiment of speed loop is designed, and the mathematical model of speed closed loop is obtained. In engineering practice, the control mode of position loop is often set to proportional (Proportion) control. In order to obtain good dynamic characteristics of the controlled system, it is necessary to optimize the control parameter P. In this paper, the iterative feedback tuning method will be used to adjust the control parameters. In the application process of iterative learning control, the error converges first, then diverges, and then converges. This paper analyzes this phenomenon from the perspective of time domain and frequency domain, and draws the conclusion that the accumulation of high frequency components in the error spectrum eventually leads to overshoot. In order to solve this problem, it is necessary to design a correlation filter to filter out the high frequency components of the error. However, the common filter has phase difference, so we introduce zero-phase filter to complete the filtering. The advance compensation of the error can make the error converge to a smaller value more quickly. Both zero-phase filtering and error advance compensation are completed in the offline state between the two iterations. The optimal number of advance compensation steps and the corresponding maximum cutoff frequency need to be determined according to the simulation experiments designed by the algorithm. According to the cause analysis of overshoot phenomenon and the related filtering and compensation mechanism, the simulation experiment is carried out, and the expected effect is obtained, and the error realizes monotone convergence. Based on the theoretical derivation and simulation experiment, a concrete NC machine tool experiment is designed. The error converges monotonously, and the final error can converge to less than 1 渭 m. The effect of zero phase filtering and error advance compensation is verified. The spectrum analysis of the error in the iterative operation process is carried out, so that the effect of zero-phase filtering can be intuitively reflected.
【学位授予单位】：哈尔滨工业大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP273
【参考文献】
相关期刊论文 前7条
1 孙明轩;毕宏博;周国良;王惠峰;;反馈辅助PD型迭代学习控制:初值问题及修正策略[J];自动化学报;2015年01期
2 曹伟;孙明;;时变离散切换系统的迭代学习控制[J];物理学报;2014年02期
3 楼晓春;陈华凌;;柔性机械臂的迭代学习控制研究[J];现代制造工程;2010年06期
4 李闯;王福利;董召然;陈鹃;;注塑机注射速度的零相位滤波迭代学习控制[J];信息与控制;2009年03期
5 施丽婷;黄筱调;杨勇;;数控交流伺服系统三环整定及应用[J];南京工业大学学报(自然科学版);2006年04期
6 许建新,侯忠生;学习控制的现状与展望[J];自动化学报;2005年06期
7 傅新,邹俊,杨华勇;迭代反馈整定在电液伺服系统中的应用[J];农业机械学报;2005年10期
相关会议论文 前1条
1 王卫红;侯忠生;金尚泰;;迭代反馈整定方法综述[A];第二十六届中国控制会议论文集[C];2007年
，本文编号：2483517
本文链接：https://www.wllwen.com/kejilunwen/zidonghuakongzhilunwen/2483517.html
