拼接主镜驱动控制技术研究

发布时间：2018-07-15 11:50

【摘要】：对分块拼接主镜的主动控制可以有效的降低拼接误差、提高成像质量。由于主镜质量较重,谐振频率较低,低拼接主镜主动控制的带宽一般在1 Hz以下,但仍能满足对温度、重力等低频扰动的校正需求。KECK、TMT和E-ELT等拼接主镜结构均采用边缘传感器测量相邻子镜的拼接误差,利用每块子镜的三个驱动器实现误差校正。但是边缘传感器的检测数据会随着时间和环境变化等产生漂移,导致不可接受的拼接误差。因此,KECK望远镜会根据观测任务需求,先对不同天顶位置主镜拼接控制量进行标定,实际工作时根据查表法进行开环控制。同时,为减小边缘传感器数据漂移的影响,每过一段时间对边缘传感器进行繁琐的标定。本文尝试将传感器标定值作为反馈,避免使用昂贵且会发生漂移的边缘传感器,以实现对主镜拼接误差的实时闭环校正。为实现对主镜静态拼接误差的高效校正和动态误差的实时校正,要求控制系统有足够的静态增益和一定的控制带宽。因此本文首先研究了被控对象影响函数模型,并根据这一模型进行控制器的设计和实验研究。被控对象模型包含影响函数和延时。影响函数表征了驱动器位移与分块主镜表面给定位置位移之间的关系,是一个与控制带宽无关的关系矩阵,在KECK中称之为面型控制方程,为了方便区分本文称之为影响函数。被控对象模型除影响函数之外还包含一个延时环节,延时是指驱动器发出控制命令到传感器发生相应改变的时间。带来延时的因素有很多,比如计算用时、传感器、采样器、DA转换和AD转换,在本文中实测结果是1个采样周期。控制器包含控制矩阵和单变量控制器。拼接主镜主动控制系统是一个多变量控制系统,且变量之间存在耦合作用。为了实现拼接主镜主动控制,首先利用控制矩阵消除变量之间的相互耦合,然后对解耦之后的系统按照经典控制理论进行控制器设计。影响函数与控制矩阵互为逆矩阵,控制矩阵在多变量控制器设计中起着至关重要的作用。影响函数一般不存在逆矩阵,但通过最小二乘法和奇异值分解法可获得影响函数广义逆。多变量控制器主要基于最小二乘逆解耦控制和SVD分解控制的进行设计,根据Simulink仿真结果,两者都可以实现被控对象的稳定控制,且对影响函数元素变化不敏感。解耦之后得到的单变量系统是延时环节,延时环节可采用积分器进行控制,积分控制器设计必须兼顾相位裕度、误差带宽、灵敏度函数增益和阶跃响应调节时间。将设计完成的解耦控制器应用在主动共焦共相实验系统中,顺利实现了拼接主镜主动共焦。对系统的分析发现,系统倾斜误差校正带宽达到0.34 Hz,实现了校正低频扰动的目的;阶跃响应调节时间2.6 s,超调量约为7%,实现了静态拼接误差的快速校正。需要指出的是由于检测等方面的原因,还未实现主动共相调整。
[Abstract]:The active control of segmented primary mirror can effectively reduce the splicing error and improve the imaging quality. Because of the heavy quality of primary mirror and low resonant frequency, the bandwidth of active control of low splicing primary mirror is generally less than 1 Hz, but it can still satisfy the temperature of the primary mirror. Correction requirements for gravity isofrequency disturbances. The spliced primary mirror structures such as KECKT TMT and E-ELT all use edge sensors to measure the splicing errors of adjacent sub-mirrors and use the three drivers of each sub-mirror to achieve error correction. However, the detection data of edge sensor will drift with time and environment, resulting in unacceptable splicing error. Therefore, according to the requirements of the observation mission, the KECK telescope will first calibrate the control quantities of the primary mirror splicing at different zenith positions, and carry out open-loop control according to the look-up table method in practice. At the same time, in order to reduce the influence of edge sensor data drift, the edge sensor is calibrated every time. This paper attempts to use the calibration value of the sensor as feedback to avoid using the expensive edge sensor which will drift in order to realize the real-time closed-loop correction of the primary mirror splicing error. In order to correct the static splicing error of primary mirror efficiently and to correct the dynamic error in real time, it is required that the control system has enough static gain and certain control bandwidth. In this paper, the influence function model of the controlled object is studied, and the controller design and experimental research are carried out according to the model. The controlled object model includes influence function and delay. The influence function represents the relationship between the displacement of the driver and the displacement of the given position on the surface of the block primary mirror, which is a matrix independent of the control bandwidth. It is called the surface control equation in KECK, and in order to distinguish the influence function in this paper. In addition to the influence function, the controlled object model also includes a delay link, which refers to the time when the driver sends out the control command and the sensor changes accordingly. There are many factors that bring delay, such as computing time, sensor, sampler DA conversion and AD conversion. In this paper, the measured results are a sampling period. The controller consists of a control matrix and a single variable controller. The active control system of splicing primary mirror is a multivariable control system, and there is coupling between variables. In order to realize the active control of spliced primary mirror, the control matrix is firstly used to eliminate the mutual coupling between variables, and then the controller is designed according to the classical control theory for the decoupled system. The control matrix plays an important role in the design of multivariable controller. Generally speaking, there is no inverse matrix in the influence function, but the generalized inverse of the influence function can be obtained by the least square method and the singular value decomposition method. The multivariable controller is mainly based on the least-squares inverse decoupling control and SVD decomposition control. According to the Simulink simulation results, both of them can realize the stable control of the controlled object, and are not sensitive to the change of the influencing function elements. After decoupling, the single variable system is a delay link, which can be controlled by an integrator. The design of the integral controller must take into account the phase margin, error bandwidth, sensitivity function gain and step response adjustment time. The designed decoupling controller is applied to the active confocal co-phase experimental system and the active confocal joint primary mirror is successfully realized. Through the analysis of the system, it is found that the correction bandwidth of the system tilting error reaches 0.34 Hz, the aim of correcting the low frequency disturbance is realized, and the step response adjusting time is 2.6 s, the overshoot is about 7, and the fast correction of static splicing error is realized. It should be pointed out that due to detection and other reasons, active phase adjustment has not been achieved.
【学位授予单位】：中国科学院大学(中国科学院光电技术研究所)
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP273

【参考文献】