节点文献
基于FTS的微结构表面超精密车削控制系统及实验研究
Research on the Control System and Experimental Study of Ultra-Precision Diamond Turning of Microstructured Surfaces Based on FTS
【作者】 王晓慧;
【导师】 孙涛;
【作者基本信息】 哈尔滨工业大学 , 机械制造及其自动化, 2011, 博士
【摘要】 微结构表面是指具有特定微小拓扑形状的功能表面,由于其独特的光学特性、粘附性、摩擦性、耐腐蚀性等,在民用和军用工业领域中有着广泛的应用。利用快速伺服刀架(FTS)作为精密微位移加工模块来车削加工微结构表面的方法已经成为微结构表面切削加工的一种主流技术。目前,世界发达国家已经研发了基于FTS的微结构表面切削加工的超精密加工设备及其多轴数控系统,并成功的实现了高精度微结构表面的加工。国内对微结构表面切削加工的研究仅仅处于起步阶段,无论是加工设备还是加工质量与国外相比还具有相当大的差距。基于此,本文搭建了可用于复杂微结构表面加工的多轴数控系统,并针对微结构表面切削加工过程中的控制与优化问题进行了相应的研究。建立多轴联动数控系统是实现非回转对称微结构表面加工的前提条件。为了满足加工实时性的要求,搭建的数控系统将FTS模块作为一个加工轴来对待,这样就避免了为FTS设置独立的控制模块所带来的其他加工轴与之通讯的问题。由于在非回转对称微结构表面的加工中FTS的输出位移量是由主轴的角度位置θ和X轴方向的刀具位置共同决定的,本文利用UMAC控制器的时基触发控制方法实现FTS进给与主轴转角和X轴进给的精确同步,实现非回转对称微结构表面的加工。建立面向微结构加工的数控系统人机界面,及能够实现典型微结构表面加工的自动编程系统。作为FTS的驱动元部件,压电陶瓷本身所固有的迟滞、蠕变等非线性特点不但会降低FTS系统的控制精度,而且可能造成系统失稳。因此,为了减小迟滞非线性的影响,提高FTS的跟踪控制精度,本文利用拓展输入空间法建立FTS迟滞系统的RBF神经网络模型,并在其逆模型基础上实现了FTS的闭环控制。由于理论上神经网络只能对一一映射或多对一映射建模,不能辨识FTS系统的迟滞特性这类多值映射的非线性现象,因此引入迟滞算子的概念,将迟滞算子的输出与系统的输入一起作为神经网络的输入向量,实现迟滞系统由多值映射到单值映射的转换。由于实际加工过程中使用的UMAC控制器将快速伺服刀架(FTS)也作为一个加工轴来对待,因此,利用UMAC控制器底层伺服控制开放性这一特点,编写自定义伺服算法代替控制器内建的PID算法,达到了对FTS精确控制的目的。为了提高加工工件的表面精度,误差补偿方法是最为经济有效地方法之一。由于在微结构表面的超精密车削加工过程中,切削深度是随着微结构表面的轮廓不断变化的,不同于普通的平面切削加工。因此,针对微结构表面切削加工的特点,我们提出了基于最小二乘支持向量机(LS-SVM)的误差补偿方法。通过分析在不同加工参数下得到的微结构表面误差轮廓,利用LS-SVM在小样本空间下的回归功能建立微结构表面切削加工过程的误差模型,并利用FTS作为误差补偿机构用以补偿在微结构表面切削过程中各种系统误差因素对加工表面轮廓精度的影响。通过回转对称正弦波微结构表面的误差补偿加工实验来证明该误差补偿方法的有效性。应用搭建的多轴联动控制系统进行了相应的加工实验研究,成功实现了五波瓣微结构表面、正弦网格微结构表面等非回转对称微结构表面的车削加工。利用二维离散Fourier变换(2D-DFT)对已加工的正弦网格表面进行评价,分析了其表面误差频率成分及成因,并提出了相应的改进策略。
【Abstract】 Micro-structured surfaces which have particular micro-topological structures are widely used in military and civilian industries applications because of their special properties, such as optical, conglutinative, frictional characteristics. Precision diamond turning based on a fast tool servo as a micro-displacement module is a very popular micro-structured surface manufacturing technology. Nowadays, some developed countries have successfully fabricated precision micro-structured surfaces applying their precision equipment and numerical control multi-axis systems. However, we are on our first stage of researching the diamond turning of micro-structured surfaces, and there is still a large gap between our country and developed countries in the research of manufacturing equipment and fabricating quality. Based on this, a CNC system for machining of complex micro-structured surface is established and some relevant works of control and optimization during the process of micro-structured turning have been studied in this thesis.Multi-axis numerical control system is a precondition for obtaining non-symmetric micro-structured surfaces. In order to satisfy the real-time requirement, the fast tool servo system is taken as a universal machining axis so as to avoid the communication problems between the FTS and other machining axis caused by establishing a dependent control module for FTS. For non-symmetric micro-structured surface turning, the displacement of FTS is determined by the angle position of the spindle and the position of X-slide. In order to generate free form micro-structured surfaces, the displacement of FTS has to synchronize with the rotation of the spindle and the movement of the X-slide. By applying the UMAC’s time-base trigger control method, the synchronization has been achieved and the non-symmetric micro-structured surfaces have been successfully fabricated. The numerical control system suited for the complex micro-structured surfaces turning and the auto programming system are also developed.Piezoelectric actuator is used as universal driving base of the FTS because of its high resolution and high stiffness. However, piezoelectric actuator exhibits hysteresis in their response to an applied electric field, which not only considerably degrades the dynamic performance of the system but also leads to inaccuracy of the FTS system. A hysteresis model using RBF neural networks of FTS is established based on expanded input space method, and closed loop control of FTS is applied using corresponding inverse model. However, it is known that neural networks can only be available for the approximation of the continuous systems with one-to-one or multiple-to-one mappings. It is unable to use the neural networks to directly identify the model of systems with multi-valued mapping such as FTS hysteresis. In this paper, a novel hysteresis operator is proposed to construct an expanded input space so as to transform the multi-valued mapping into a one-to-one mapping which enables neural networks to approximate the behavior of hysteresis of FTS. In order to achieve better tracking performance of FTS, which is treated as a universal machining axis, user defined servo algorithm is programmed to take the place of PID algorithm embedded in UMAC controller.Error compensation could be the one of the most effective and economic strategies to improve the quality of machined micro-structured surfaces. Error compensation method applying the least square support vector machine (LS-SVM) is proposed according to the particularity of micro-structured surfaces turning with varing depth of cut which is quite different from general face cutting. In order to reduce the influence of systematic error factors to the contour accuracy, a least square support vector regression model of micro-structured surfaces turning errors is established by analysing the contour errors of micro-structured surfaces machined with different cutting parameters. The compensation experiment has been excuted and indicated the effectiveness of the proposed compensation method.The machinging experiments were carried out with the established multi-axis numerical control system and Non-rotation symmetric surface structures—five bode and sinusoidal grid microstructured surfaces were successfully manufactured by the designed machining system. An evaluation technique based on the two-dimensional discrete Fourier transform (2D-DFT) of the sinusoidal grid interference microscope images was developed to evaluate the fabricated micro-structured surfaces effectively. According to the frequency components of the spectrum obtained by the 2D-DFT of the image, error sources were analysed and corresponding amendment strategies were proposed.