-
激光干涉测量转轴径向运动误差的原理如图1所示,主要包括测量单元和移动单元两部分,其中测量单元包含单频激光器和偏振分光棱镜等光电器件,移动单元包含伺服转轴和角锥棱镜等。
激光干涉的测量臂与待测转轴径向运动误差方向平行,伺服转轴同轴安装于待测转轴上方,角锥棱镜安装于伺服转轴上方。伺服转轴相对于待测转轴反向旋转,从而保证入射至角锥棱镜的激光光线始终能够返回测量单元,并与参考臂返回的激光光线在偏振分光棱镜PBS1处汇合之后经非偏振分光棱镜NPBS等光学器件在四个光电探测器上产生四路相位依次相差90°的干涉信号,如图2(a)所示。然后将相位相差180°的两路信号分别进行电路差分,消除共模噪声,得到一组包含相位和位移方向信息的正交信号,如图2(b)所示。
设待测转轴的径向运动误差为δx,则在干涉测量的两臂之间产生的光程差为2δx,干涉信号的周期变化次数N为:
$$ N = \dfrac{{{\text{2}}{\delta _x}}}{\lambda } = \dfrac{\varphi }{{{\text{2}}\pi }} $$ (1) 式中:λ为空气中激光的中心波长;φ为干涉信号的总相位变化值。可见,要实现转轴径向运动误差的高精度测量,需要对干涉信号的相位值进行高精度解算。
-
激光干涉测量是以微米量级的激光波长为测量基准,为了进一步达到纳米级的测量精度,要对干涉信号进行相位细分。CORDIC算法通过角度旋转迭代对相位信号进行反正切计算,如图3所示,该算法收敛速度快、精度高,且避免了复杂的乘除、开方等运算,适用于嵌入式处理系统。
CORDIC算法迭代公式为:
$$ \left\{ {\begin{array}{*{20}{c}} {{x_{i + 1}} = {x_i} - {d_i}({2^{ - i}}{y_i})} \\ {{y_{i + 1}} = {y_i} - {d_i}({2^{ - i}}{x_i})} \\ {z{}_{i + 1} = {z_i} - {d_i}{\theta _i}} \end{array}} \right. $$ (2) 式中:i=0,1,···,n;θi=arctan (2−i);令z0=0,zn即为所求相位;di为方向判断因子,取值±1。由于计算相位的过程是有限旋转迭代,必然导致计算结果θ与真值θ0之间存在残余角度误差Δθ:
$$ \Delta \theta = {\theta _0} - \theta = {\theta _0} - \displaystyle\sum\limits_{i = 0}^n {{\theta _i}} $$ (3) 可见,迭代次数n越大,残余角度误差越小。针对任意两个时刻之间的一段干涉信号相位解算,为了保证精度并减少运算量,文中研究提出将CORDIC算法仅用于首尾不足1/8周期的信号相位处理,得到高精度的首尾相位小数计数值n1和n2,而对中间段的信号则采用八细分算法。八细分算法是通过比较正交信号函数值的符号和绝对值将一个周期均匀分成8部分,如图4所示,只需较低的运算量即可实现信号1/8周期整数的计数N1/8。
因此,信号的整体波形细分计数原理如图5所示。根据公式(1),转轴径向运动误差可表示为:
$$ {\delta _x} = \dfrac{\lambda }{2}N = \dfrac{\lambda }{2}\left( {\dfrac{{{N_{{1 \mathord{\left/ {\vphantom {1 8}} \right. } 8}}}}}{8} + {n_1} + {n_2}} \right) $$ (4) 综合考虑计算量和测量分辨率的要求,CORDIC算法迭代次数选择10,对应理论分辨率约为0.97 nm。
-
正交干涉信号的质量是影响上述相位细分算法精确性和稳定性的重要因素。然而,由于光源不稳定性、光电器件自身特性不理想以及环境参数变化等原因,正交干涉信号不可避免的存在非线性误差,包括直流偏差p、q,交流幅值不等Ax、Ay和非正交误差α,以上简称三差。因此,实际正交干涉信号可表示为:
$$ \left\{ {\begin{array}{*{20}{c}} {{U_y} = {A_y}{\text{sin(}}\theta {\text{)}} + p } \\ {{U_x} = {A_x}{\text{cos(}}\theta + \alpha {\text{)}} + q} \end{array}} \right. $$ (5) 通过准确求解正交信号的峰谷值可以对直流偏差和交流幅值不等误差进行修正,具体如公式(6)和(7):
$$ \left\{ {\begin{array}{*{20}{c}} {{A_y} = \dfrac{{\text{1}}}{{\text{2}}}\left( {{U_{y\_\max}} + {U_{y\_\min}}} \right)} \\ {{A_x} = \dfrac{{\text{1}}}{{\text{2}}}\left( {{U_{x\_\max}} + {U_{x\_\min}}} \right)} \end{array}} \right. $$ (6) $$ \left\{ {\begin{array}{*{20}{c}} {p = \dfrac{1}{2}\left( {{U_{y\_\max }} + {U_{y\_\min }}} \right)} \\ {q = \dfrac{1}{2}\left( {{U_{x\_\max }} + {U_{x\_\min }}} \right)} \end{array}} \right. $$ (7) 式中:Uy_max和Ux_max为两信号的峰值;Uy_min和Ux_min为两信号的谷值。
对非正交误差的修正可通过矢量运算的形式将两信号转化为相位一致的信号,从而无需解算非正交误差,即可完成修正。公式(8)中,U1和U2是消除直流偏差和幅值不等误差后的两信号,β为修正后产生的新相位。非正交误差的修正将使两信号再次出现幅值不等,需要重新归一化处理。
$$ \left\{\begin{array}{l} U_y{ }^{\prime}=U_1+U_2=A \cdot \sqrt{2+2 \cos \alpha} \cdot \sin (\theta+\beta) \\ U_x^{\prime}=U_1-U_2=A \cdot[-\sqrt{2+2 \cos \alpha} \cdot \sin (\theta+\beta)] \end{array}\right. $$ (8) 由上述修正原理可知,信号峰谷值的准确求解是三差信息提取和非线性误差修正的前提和关键。然而,由于伺服转轴和待测转轴的自身转速不匀、伺服响应延迟或跟踪不稳等原因,实验发现转轴径向运动误差的干涉测量信号中存在大量随机微幅抖动,传统的信号峰谷值提取方法准确度差,三差信息提取存在较大偏差,无法实现非线性误差的有效修正。
图6所示为信号相位抖动的仿真示意图。传统方法利用信号零点或极值点位置进行信号截取并提取峰谷值,仅适用于图6中第3段所示的不含相位抖动的信号,而针对第1和第2段包含位置与幅度随机抖动的转轴径向运动误差干涉测量信号不能适用。特别是当相位抖动位于极值点或零点位置附近,单凭零点或极值点截取信号会将抖动识别为一个修正周期,造成峰谷值提取错误。此外,将极值点作为修正周期截取的标准,会增大硬件电路运算压力。
文中提出零值截取-阈值判定的新方法,采用自适应窗进行取样,将三次有效零点位置作为一个修正周期截取标准,再配合阈值判定,可以准确提取峰谷值,且运算量小。具体如下:
以蓝色正弦信号为例(以任意一只正交信号中的零值点位置作为周期划分标准不影响修正效果),信号经过三次零点为初始取样长度,将该信号段内峰值、谷值与设定阈值进行比较。若峰值和谷值均超过阈值,说明该信号段包含了有效的三差信息;未超过阈值,则说明相位在持续抖动,需要扩大截取范围,寻找下一个零值点,直至信号段内峰谷值超过设定阈值。阈值的设定与激光功率、光电探测器响应以及后续信号放大倍率相关,与被测轴无关,考虑激光功率存在零漂,可将阈值设定为理论峰谷值的0.8倍左右。
图6截取信号段a-b-c,零值点b、c处有明显的相位抖动,实际信号并未达到谷值位置,此时峰谷值有一项未超过阈值,需要扩大信号段截取范围,当信号截取范围扩大至a-b-c-d-e,此时,峰谷值均超过阈值,该信号段为包含准确峰谷值的有效修正周期。将此时的峰谷值带入三差修正算法,能够实现非线性误差的有效修正。
Signal processing method for measuring the radial motion error of rotary axis by interferometry
-
摘要: 转轴径向运动误差是转轴的重要误差项之一,严重影响数控机床等转轴相关设备的精度和性能。利用激光干涉配合伺服转轴可以实现转轴径向运动误差的测量,但由于待测转轴和伺服转轴的自身转速不匀、伺服响应延迟、跟踪不稳等原因,干涉测量信号存在持续微幅相位抖动,进而造成非线性误差难以有效修正,相位解算精度不高,测量误差大。针对这一问题,提出一种零值截取-阈值判定的干涉信号处理方法,成功消除了相位抖动的影响。设计并搭建了一套转轴径向运动误差的激光干涉测量装置,针对实测信号的处理结果表明,相比于传统修正方法,文中提出的修正方法使得转轴径向运动误差干涉测量信号解算的重复性由4.8 μm减小到0.2 μm,与标准仪器的对比误差由3.5 μm降为2 μm。Abstract: The radial motion error is one of the most important errors of rotary axis, which seriously affects the precision and performance of CNC machine tools. Using laser interference with reference rotary axis can achieve measurement of radial motion error. Neither target or reference rotary axis is ideal, there are some restrictions like uneven speed, response delay and unstable tracking. These restrictions cause phase jitter and make the non-linear correction difficult. It will affect the phase calculation accuracy of the interference signal. In order to eliminate the influence of phase jitters, a type of signal processing method is proposed. The paper designs and builds a set of laser interference measuring device for measuring the radial motion error. Compared to the traditional correction method, the proposed method reduces the measurement repeatability from 4.8 μm to 0.2 μm, and the error compared with the standard instrument is decreased from 3.5 μm to 2 μm.
-
Key words:
- rotary axis /
- radial motion error /
- phase jitters /
- on-linear correction
-
[1] Soichi Ibaraki, Wolfgang Knapp. Indirect measurement of volumetric accuracy for three-axis and five-axis machine tools: A review [J]. International Journal of Automation Technology, 2012, 6(2): 110-124. [2] Yang J, Feng Q, Li J. Review on multi-degree-of-freedom motion error measurement methods for rotary-axis laser [J]. Laser & Optoelectronics Progress, 2016, 53(9): 090003. (in Chinese) [3] Lou Z, Hao X, Liu L, et al. spindle radial motion error measurement using a circular grating and a autocollimator [J]. Optics and Precision ngineering, 2019, 27(9): 2053-2061. (in Chinese) doi: 10.3788/OPE.20192709.2053 [4] Huang B. Research on the spindle radial motion error of the numerical machine tools based on laser displacement sensor[D]. Hangzhou: Zhejiang University, 2008. (in Chinese) [5] Sun C. Research on multi-degree-of-freedom measurement method for precision engineering[D]. Changchun: University of Chinese Academy of Sciences (Changchun Institute of Optics), 2021. (in Chinese) [6] Lv Y, Feng Q, Liu L, et al. Six-degree-of-freedom measurement method based on multiple collimated beams [J]. Infrared and laser Engineering, 2014, 43(11): 3597-3602. (in Chinese) [7] Liang R, Li W, Wang Z, et al. A method to decouple the geometric errors for rotary axis in a five-axis CNC machine [J]. Measurement Science and Technology, 2020, 31(8): 085007. doi: 10.1088/1361-6501/ab7ded [8] Li Q, Wang W, Zhang J, et al. All position-dependent geometric error identification for rotary axes of five-axis machine tool using double ball bar [J]. The International Journal of Advanced Manufacturing Technology, 2020, 110(5): 1351-1366. [9] Li F, Dong H, Wang J, et al. Error analysis of rotation positioning accuracy measured by laser tracker [J]. Mechanical and Electrical Equipment, 2021, 38(3): 59-63. (in Chinese) [10] Bringmann B, Besuchet J P, Rohr L. Systematic evaluation of calibration methods [J]. CIRP Annals, 2008, 57(1): 529-532. doi: 10.1016/j.cirp.2008.03.114 [11] Xia H, Hu M, Zhang X. High precision of quadrature signals for homodyne interferometer [J]. Optics and Precision Engineering, 2017, 25(9): 2309-2316. (in Chinese) doi: 10.3788/OPE.20172509.2309 [12] Li L. Application of laser interferometer in XY table measurement [J]. Equipment for Electronic Products Manufacturing, 2020, 49(5): 37-41. (in Chinese) [13] Ping S, Fu Y, Zhang F, et al. Study on measurement method of mirror reflection laser tracking interferometric length measurement [J]. Infrared and Laser Engineering, 2021, 50(12): 20210624. (in Chinese) doi: 10.3788/IRLA20210624 [14] Li J, Feng Q, Bao C, et al. Method for simultaneously and directly measuring all six-DOF motion errors of a rotary axis [J]. Chinese Optics Letters, 2019, 17(1): 47-51. [15] Xing J, Zhao G. Research on compensation of laser interferometry [J]. Laser Journal, 2020, 41(7): 63-66. (in Chinese)