【作者】 钟俊；

【导师】 竺长安；

【作者基本信息】 中国科学技术大学 ， 精密仪器及机械， 2011， 博士

【摘要】 近年来,随着工业领域如微型机械制造、超精密测量、超精密加工、集成电路制造、生物工程、医疗科学、光纤对接和机器人系统等众多领域的不断发展,与之相关的设备对定位行程和定位精度的要求越来越高,因此宏观尺度的纳米级定位技术一直是精密工程领域的研究热点之一。同时,随着各项大型天文物理项目的不断开展,对大口径高精度的衍射光栅提出了越来越高的要求。因此本论文以衍射光栅刻划机的分度定位系统作为实验和应用对象,对其中的核心技术——宏观尺度的纳米级定位控制技术中的系统构建、定位系统建模及特性分析、定位误差补偿控制策略及宏/微两级定位方式等关键问题进行了研究。通过对衍射光栅刻划机的工作方式、系统构成和各种宏观尺度纳米级定位系统构成方案的对比分析,确定了以直流无刷力矩电机加蜗杆蜗轮、丝杆螺母机构作为宏动台、压电陶瓷加弹簧钢片作为微动台、双频激光干涉仪作为位置测量系统全闭环的宏/微两级驱动方式的宏观尺度纳米级分度定位系统。在PC机上通过VC对刻划机的分度定位系统和刻划系统进行了集成。实现了对中国科学院长春光学精密机械与物理研究所2号衍射光栅刻划机的控制。在实际光栅刻划的过程中,其分度定位系统实现了在300mm行程内,平均稳态3σ定位误差小于7nm的定位精度。在宏/微两级驱动定位系统的系统建模研究中,分析对比了常用的“白箱建模”、“黑箱建模”和“灰箱建模”三种建模方法的优缺点,并确定采用“灰箱建模”方法对定位系统进行建模。通过机理建模,根据宏、微工作台的链接方式建立了不同驱动方式(宏驱动单输入-双输出、微驱动单输入-双输出和宏/微驱动双输入-双输出)的整体动态模型,并通过系统辨识实验,对系统中的未知参数进行了估计,获得了与实际系统较接近的系统模型,根据灰箱建模得到的系统模型对系统的动态特性进行了分析,并提出了系统改进方向。通过对传统PID控制原理及实验结果的分析,阐明了其不适于宏观尺度纳米级定位系统控制的原因,在阐述了神经网络基本原理的基础上,阐明了使用神经网络PID控制作为宏观尺度的纳米级定位系统控制方法的优点,通过对单神经元PID控制和BP神经网络PID控制两种神经网络控制算法的原理分析及在“灰箱建模”得到的系统模型进行的控制仿真,对这两种PID控制算法的控制性能进行了对比分析。最后,根据本论文所研究的衍射光栅刻划机的机械结构及其分度系统的定位精度要求,搭建了实验平台,对定位系统的几个关键的驱动和传感器件进行了选择;在此基础上,采用虚拟硬件在环的方法对光栅刻划机的计算机控制软件进行了仿真开发,利用“灰箱建模”得到的控制系统模型在完全虚拟的环境下编程实现了刻划机的定位控制流程和刻划流程。通过对单神经元PID和BP神经网络PID两种算法的阶跃和步进定位实验,对上述两种控制算法的研究和仿真进行了验证,并对控制算法的快速性、定位精度等进行了对比分析得出:BP神经网络学习收敛速度更快,能够更快速地适应系统的变化,具有较好的快速性,能够获得更高的定位精度,但是系统超调较大。然后,根据衍射光栅刻划原理实现了间歇式定位和宏动台连续-微动台间歇两种定位方式的光栅刻划机的刻划控制,并进行了实际刻划实验,通过对实验结果的处理分析可以得出,目前长春光机所2号衍射光栅刻划机的宏定位系统存在爬行现象,因此采用间歇式定位方式和快速性较好的BP神经网络PID控制算法可以获得几种方法中最高的定位精度(7nm以下),而宏动台连续-微动台间歇的定位方式的精度较差。通过微定位实验和实际刻划实验验证了本论文研究工作的正确性。更多还原

【Abstract】 Recently,asthedevelopmentofnumerousindustrialareassuchasmicro-machines,ultra-precision measurement, ultra-precision machining, integrated circuit manufactur-ing, biotechnology, medical science, optical docking and robotics, the equipment withlarger positioning range and higher positioning accuracy are highly required, so themacroscopic-scale nano-precision positioning technology has been a hot in engineeringresearch field. Meanwhile, as the development of large-scale astrophysics project, therequirement on high-precision diffraction gratings in large diameter grows. Therefore,by taking the indexing and positioning system in dissertation diffraction grating rulingmachine as the research object, this dissertation investigated the system construction,system modeling and characterization, positioning error compensation control strategyand macro/micro-dual positioning in macroscopic-scale and nano-positioning technol-ogy.By comparing the operating mode and system component of diffraction gratingruling engine with that of other macroscopic-scale and nano-positioning systems, aclosed loop macroscopic-scale nano-positioning indexing and positioning system withmacro/micro dual-drive is established, with the DC brushless torque motor plus wormgear and screw nut mechanism as macro blank, with the piezoelectric ceramics plus s-teel spring asmicroblank, with thedual-frequency laser interferometerasmeasurementequipment. Through simulating the indexing positioning and ruling systems of the d-iffraction grating ruling engine on PC via VC, effective control on diffraction gratingruling engine No. 2 in Changchun Institute of Optics, Fine Mechanics and Physics ofChinese Academy of Sciences is realized. In practice, the average 3σpositioning errorof the indexing positioning system is less than 7nm within 300m positioning range.Inthesystemmodelingofthemacro/microdual-drivepositioningsystem, bycom-paring the advantages and shortcomings of the commonly used”white box modeling”,”black box modeling”and”gray box modeling”, the”gray box modeling”is adopted.According to the link mode of the macro and micro worktable, overall dynamic modelof different driving modes (Macro drive single input - dual output, Micro drive singleinput - dual output, and macro / micro drive dual input - dual output)are establishedby mechanism modeling. Through system identification experiments, unknown param-eters in systems are estimated and system model close to the actual system is obtained.Dynamiccharacterofthesystemisanalyzedbased on”graybox modeling”, andfurther improving methods are proposed.ThroughanalyzingthetraditionalPIDcontroltheoryandexperimentalresults, thisdissertation stated it is not suitable for macro-scale and nano-positioning system controland the reasons are given. Based on the principles of neural network, the advantage ofusing neural network PID control in macro-scale nano-positioning system is demon-strated. By analyzing control algorithm of the single neuron PID control and BP neuralnetwork PID control, together with the simulation of the system model based on”graybox modeling”, the performance of the two types of PID control algorithm were com-pared.Finally, according to the positioning accuracy demand of the mechanical structureandindexingsystemofdiffractiongratingrulingengine, severalkeydriversandsensorsof the positioning system were selected and the experiment platform was constructed.Based on this, the computer-controlled software of the grating ruling engine was de-veloped with the virtual hardware in the loop method, and the positioning process andruling process of the ruling engine were realized in a fully virtual environment usingcontrol system model based on”gray box modeling”. Through the step positioningexperiments on the single neuron PID and BP neural network PID control algorithm,the two control algorithm were simulated and verified, and the rapidity and accuracyof the control algorithms was compared. It is obtained that BP neural network controlconverges faster, is more adaptable to changes in the system, works faster, can achievehigher accuracy, but has larger overshoot. Then, based on the working principle ofdiffraction grating ruling engine, intermittent positioning and macro worktable contin-uous mode-micro worktable intermittent mode positioning control of the grating rulingengine were realized. Ruling experiment were conducted and the analysis on the result-s show that the macro positioning system of the diffraction grating ruling engine No.2 in Changchun Institute of Optics, Fine Mechanics and Physics of Chinese Academyof Sciences has crawl, therefore, by using intermittent positioning and fast BP neuralnetwork PID control algorithm, high accuracy (7nm or less)can be achieved, whilethe accuracy performance of the macro worktable continuous mode-micro worktableintermittent mode positioning is poor. The correctness of the work in the dissertation isverified by micro-positioning experiments and practical ruling experiments.更多还原