光场式直线时栅高精度动态测量方法研究

发布时间：2018-03-15 04:20

本文选题：光场式直线时栅　切入点：动态测量　出处：《重庆理工大学》2017年硕士论文　论文类型：学位论文

【摘要】：精密位移测量技术是一个国家制造业走向高、精、尖的基础保障。随着科学技术的快速发展,自动化设备对精度要求的提高,精密动态测量技术在现代科学技术领域已占有重要的地位。时栅是中国人自主发明的全新原理的位移传感器,其中光场式直线时栅是时栅的另一个分支研究方向。目前,光场式直线时栅在静态的测量实验中,精度为±2μm,速度在0~5mm/s的动态测量实验条件下,精度为±8μm,但在速度大于5mm/s的实验中,测量精度有明显的下降,具体表现在测量的位移量滞后于真实的位移量,即“时-空”不同步,并且在不同的测量速度下,分辨力发生变化,这两个问题是动态误差产生的主要原因。本文在国家自然科学基金项目“一种交变光场时空耦合的高速高精度位移传感器研究”的资助下,针对传感器的测量速度大于5mm/s动态测量中存在的两个问题,开展了高精度的动态测量方法的研究。研究目标:研究两种提高光场式直线时栅动态特性的方法,以期在5mm/s以上动态测量实验中,传感器的测量精度做到±4μm以内。首先,提出了提高光场式直线时栅动态性能的两种方法—BP神经网络预测算法、连续动态比相法,对这两种方法建立了相应的数学模型,结合时栅传感器的工作原理,阐述了两种方法的原理与可行性分析。其次,设计出了两种方法的硬件电路原理图和相应的PCB电路板,其中BP神经网络预测算法的硬件电路是以FPGA为核心,主要包括激励信号模块、双通道A/D采集模块、行波信号合成电路、UART通信模块等。连续动态比相法的硬件电路设计主要包括A/D采集模块、D/A转换模块、双向比相电路等,以FPGA作为微处理器对各模块作逻辑运算。最后,搭建了实验平台,对以上两种方法进行动态标定和实验验证,在加速度接近于0mm?,速度在9mm/s的近似匀速运动状态下,BP神经网络预测算法位移的预测误差峰-峰值为±4μm,连续动态比相法的误差峰-峰值为±3μm。在加速度大小为4mm?至5mm?时,BP神经网络预测算法的误差峰-峰值为±9μm,连续动态比相法的误差峰-峰值为±6μm。实验结果表明:两种方法都实现了预期的研究目标,但在加速度较大的测量条件下,BP神经网络预测算法与连续动态比相法的测量精度都有所下降。
[Abstract]:Precision displacement measurement technology is the basic guarantee of high, fine and sharp manufacturing industry in a country. With the rapid development of science and technology, the requirement of precision for automation equipment is improved. Precision dynamic measurement technology has played an important role in the field of modern science and technology. The time-grating is a new principle displacement sensor invented by Chinese people, among which the light-field linear time-grating is another branch of the time-grating. In the static measurement experiment, the accuracy of the light field straight line grating is 卤2 渭 m, and the accuracy is 卤8 渭 m when the velocity is 0 ~ 5 mm / s, but in the experiment with the velocity greater than 5 mm / s, the measurement accuracy is obviously decreased. The measured displacement lags behind the real displacement, that is, "time-space" is out of sync, and the resolution changes at different measuring speeds. These two problems are the main causes of the dynamic error. This paper is funded by the National Natural Science Foundation of China, a high-speed and high-precision displacement sensor with space-time coupling of alternating light fields. In view of the two problems existing in the dynamic measurement of sensor measuring speed greater than 5mm / s, the research of high precision dynamic measurement method is carried out. Objective: to study two methods to improve the dynamic characteristics of light field linear grating. It is expected that in the dynamic measurement experiment above 5 mm / s, the measurement accuracy of the sensor is within 卤4 渭 m. Firstly, two methods to improve the dynamic performance of light field straight line gate, the BP neural network prediction algorithm, the continuous dynamic phase comparison method, are proposed. The corresponding mathematical models of the two methods are established, and the principle and feasibility analysis of the two methods are expounded in combination with the working principle of the time grating sensor. Secondly, the hardware circuit schematic diagram and the corresponding PCB circuit board of the two methods are designed. The hardware circuit of BP neural network prediction algorithm is based on FPGA, including excitation signal module, dual channel A / D acquisition module. The hardware circuit design of continuous dynamic phase comparison method mainly includes A / D acquisition module / D / A conversion module, bidirectional phase comparison circuit and so on. The FPGA is used as microprocessor to perform logic operation on each module. The experimental platform is built to calibrate and verify the above two methods dynamically, and the acceleration is close to 0mm? The prediction error peak to peak value of displacement of BP neural network is 卤4 渭 m, the error peak of continuous dynamic phase comparison method is 卤3 渭 m, and the error peak of continuous dynamic phase comparison method is 卤3 渭 m under the condition of approximately uniform velocity motion of 9 mm / s, the prediction error peak to peak value of BP neural network is 卤4 渭 m, and the magnitude of acceleration is 4 mm? To 5mm? The error peak to peak value of BP neural network prediction algorithm is 卤9 渭 m, and the error peak to peak value of continuous dynamic phase comparison method is 卤6 渭 m. The experimental results show that both methods achieve the expected research goal. However, the accuracy of BP neural network prediction algorithm and continuous dynamic phase comparison method are decreased under the condition of high acceleration.
【学位授予单位】：重庆理工大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP212;TP183

【参考文献】