高精度圆光栅测角系统误差补偿技术研究

发布时间：2018-07-28 19:05

【摘要】：随着生产实践的进一步发展,对于测角系统的精度和稳定性等提出了更高的要求。基于Renishaw增量式圆光栅的高精度测角系统,本课题研究采用多读数头结构和软件补偿算法提高系统的测角精度。基于误差建模,分析多读数头结构的消除误差原理和应用场景,从原理和仿真实验角度对比分析了数值补偿算法、BP网络补偿以及基于小波神经网络(WNN)的误差回溯补偿三类补偿算法对于系统的补偿效果及其适用场景。通过圆光栅测角系统误差的建模,定量分析了系统误差、随机误差和安装误差的组成成分,推导了多读数头结构消除误差原理,定量地分析了安装倾斜和安装不对径下多读数头的误差的消除效果。分析了分段线性、三次样条插值、多项式以及谐波补偿四种数值补偿算法,并且对离散标定的测角误差进行了两次迭代补偿实验,对比分析四种算法补偿效果和适用性,其中谐波补偿测角误差补偿效果最好。分析了BP神经网络算法原理和方法,通过改变训练算法和调整网络结构改进了BP网络,并且对误差进行两次迭代实验,对比分析了其与数值补偿算法的补偿效果。为了优化系统设计,研究了基于小波神经网络的误差溯源补偿方法,并且采用小波网络方法对圆光栅测角误差进行溯源补偿,比较了其狭义和广义小波网络两种算法的补偿效果,实验验证了小波网络算法、数值补偿算法、BP网络算法三种算法的补偿有效性以及分析了它们的补偿效果和适用场合。狭义小波网络兼具神经网络和小波变换的特性,补偿效果比BP网络好;通过实验比较广义小波网络和谐波补偿算法,广义小波网络能辨识出来附加误差成分,因此达到的误差补偿效果比谐波补偿好,适合动态误差的补偿。
[Abstract]:With the further development of production practice, higher requirements for accuracy and stability of angle measuring system are put forward. Based on the high precision angle measurement system of Renishaw incremental circular grating, this paper studies how to improve the angle measurement accuracy of the system by using multi-readout head structure and software compensation algorithm. Based on error modeling, the error elimination principle and application scenario of multi-reading head structure are analyzed. From the point of view of principle and simulation experiment, the compensation effect of numerical compensation algorithm (BP) and error backtracking compensation based on wavelet neural network (WNN) for the system and its applicable scenarios are compared and analyzed. By modeling the error of circular grating angle measuring system, the components of system error, random error and installation error are quantitatively analyzed, and the principle of eliminating error by multi-reading head structure is deduced. The effect of eliminating the error of multiple reading heads under the installation of tilting and misalignment is quantitatively analyzed. Four numerical compensation algorithms, piecewise linear, cubic spline interpolation, polynomial and harmonic compensation, are analyzed, and two iterative compensation experiments are carried out for the angle measurement error of discrete calibration. The compensation effect and applicability of the four algorithms are compared and analyzed. The compensation effect of harmonic compensation for angle measurement error is the best. The principle and method of BP neural network algorithm are analyzed, the BP network is improved by changing the training algorithm and adjusting the network structure, and the error is tested twice, and the compensation effect of the algorithm is compared with that of the numerical compensation algorithm. In order to optimize the system design, the compensation method of error traceability based on wavelet neural network is studied. The compensation effects of the two algorithms are compared. The experimental results show that the compensation effectiveness of the wavelet network algorithm and the numerical compensation algorithm / BP network algorithm are verified and their compensation effect and application situation are analyzed. The narrow wavelet network has the characteristics of neural network and wavelet transform, and the compensation effect is better than that of BP network. By comparing the generalized wavelet network with harmonic compensation algorithm, the generalized wavelet network can identify the additional error component. Therefore, the error compensation effect is better than harmonic compensation, which is suitable for dynamic error compensation.
【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】：2016
【分类号】：TH741.2

【相似文献】