基于统一框架的数控机床热误差建模方法

为了克服独立筛选关键温度点再进行热误差建模破坏其内在联系从而降低热误差模型预测性能的问题,提出了一种统一框架下同时筛选关键温度点和热误差建模的方法。采用最小二乘支持向量机作为基本热误差模型,将温度点的选择状态和模型超参数作为优化变量,采用二进制鲸鱼优化算法进行寻优,并综合考虑最大化预测精度和最小化关键温度点个数设计损失函数。以一台卧式加工中心为例,进行热误差实验,利用所提方法在10折交叉验证模式下筛选出了最优关键温度点,将其个数从20减少到了3,并同时获得了模型最优超参数。最后,与传统独立方式进行了对比分析,结果表明利用所提建模方法热误差预测精度最高提高约62.8%,验证了其有效性和优越性,为后续热误差补偿实施提供了参考。     

High accurate squareness measurement squareness method for ultra-precision machine based on error separation

Traditional measurement methods of squareness for ultra-precision motion stage have many limitations, especially the errors caused by the inaccuracy of standard specimens. On the basis of error separation, this paper presents a novel method to measure squareness with an optical square brick. The angles between the guideways and the four lines of brick section are measured based on the fact that sum of interior angle of a quadrilateral is 2π, and the squareness is obtained. A squareness measurement experiment was performed on a profilometer with a modified optical square brick. Experimental results show that the squareness accuracy between X and Y axes is not influenced by the accuracy of brick, and the measurement repeatability reaches 0.22 arcsec. Finally, a verification experiment to the proposed method was carried out with a high accurate standard specimen, and the error between the two methods is 1.06 arcsec. According to the error results and simulation analysis of the measurement system, the measurement error based on error separation is 0.06 arcsec. The proposed method is able to achieve a very high accurate squareness measurement with auxiliary components of normal accuracy, and can be applied to measure the accuracy class of sub-arcsec squareness.     

基于迁移学习的异工况下机床热误差建模方法

机床热误差预测模型在不同工况下难以保持高预测精度是导致热误差实际补偿效果差的重要原因,对此本文提出一种基于迁移学习的异工况下机床热误差建模方法。首先利用核均值匹配算法获取不同工况下机床温度数据间的迁移权重,从而提出基于迁移学习的热误差建模方法;对不同工况下热误差数据进行差异显著性检验,并利用本文所提方法建立热误差预测模型,分析建模效果;然后比对分析本文所提建模方法与常用建模方法的实际预测效果,最后进行补偿验证实验以证明本文所提方法的有效性。结果表明,本文所提基于迁移学习的建模方法能够有效提升建模效果,其中迁移学习结合LASSO算法针对不同工况下热误差数据的预测精度和稳健性分别达到3.73和1.14μm,补偿后机床X/Y/Z 3个方向热误差分别保持在-2.3～3.1μm、-3.4～3.9μm和-3.3～4.6μm范围内。     

超精密五轴机床的几何误差建模和灵敏度分析

基于设计出超精密机床的目的,研究了机床的几何误差建模和误差的灵敏度分析。基于刚体运动学和齐次变换矩阵(Homogeneous Transformation Matrix,HTM)建立了RTTTR配置的超精密五轴机床的几何误差模型,模型涉及37个误差分量。分别对37个误差分量进行了几何误差的灵敏度分析,分析结果将应用于超精密五轴机床的设计与制造上。     

Thermal sensor selection for the thermal error modeling of machine tool based on the fuzzy clustering method

Thermal sensor selection is a work of great importance when modeling thermal error. The proper selection of thermal sensors and their locations may greatly improve the prediction accuracy. In this article, the fuzzy C means (FCM) clustering method and the ISODATA method are used to group the data of thermal sensors and a genetic algorithm–back propagation artificial neural network thermal model is established to testify the accuracy. A validity criterion for the FCM method is put forward to guarantee the precision of the model. Both the FCM and the ISODATA methods are effective for thermal sensor selection.     

Volumetric error modeling and sensitivity analysis for designing a five-axis ultra-precision machine tool

The five-axis machine tools are increasingly popular for meeting the demand of machining the workpiece with growing geometric complexity and high accuracy. This paper studies the volumetric error modeling and its sensitivity analysis for the purpose of machine design. The volumetric error model of a five-axis machine tool with the configuration of RTTTR is established based on rigid body kinematics and homogeneous transformation matrix, in which 37 error components are involved. The sensitivity analysis of volumetric error regarding 37 error components is carried out respectively. The analysis results are successfully used for the accuracy design and manufacture of a five-axis ultra-precision machine tool. The preliminary experiment of machining sine grid surface testifies the high accuracy and effectiveness of the designed five-axis machine tool.     

基于正则化的机床热误差自适应稳健建模算法

通过建立预测模型对机床热误差进行补偿,是有效解决热误差造成机床精度下降问题的常用方法。本文提出一种基于正则化的数控机床热误差自适应稳健建模算法,能够在建模过程中自适应选择温度敏感点(TSPs),并具有高预测精度和稳健性。首先基于结构风险最小化原则对热误差建模稳健性机理进行分析,进而利用正则化算法中LASSO解的稀疏性实现自适应TSP选择。然后基于不同实验条件的热误差数据,分析所提建模算法的预测效果,并与常用的多元线性回归、BP神经网络和岭回归算法进行比对分析。结果表明,本文所提建模算法具有最高的预测精度和稳健性,分别为5.22和1.69μm。最后,利用所建立的预测模型进行热误差补偿实验,以验证本文所提建模算法的实际补偿效果。     

Design and dynamic optimization of an ultra-precision micro grinding machine tool for flexible joint blade machining

The influences of tool setting errors on skiving accuracy are discussed in this paper. Firstly, a method for the calculation of error-free cutting edge of skiving cutter is proposed. Based on the theories of cutter enveloping gear, the gear profile deviations, which are affected by the tool position and orientation errors in skiving, are analyzed. Results show that the profile deviations are insensitive to this kind of setting error of the cutter. Then, the effects of tool eccentricity error on skiving accuracy are analyzed. Results show that the waves on the tooth flanks, which are caused by the tool eccentricity error, have obvious influences on the helix deviations and the profile deviations of the workpiece. The waves on the tooth flanks are periodic ones and can be obviously affected by the number of teeth of the cutter.     

Thermal error compensation method for machine center

High-speed machining (HSM) technology has become the most important application in metal cutting industries. However, overcome the positioning error is one of a great concern. The position error happens due to poor machine structure design and thermal expansion comes from cutting, especially when HSM is applied. In this paper, a new technique is developed to compensate for these errors. The thermal images are used to confirm dispersion of all the affect positions. PT-100 thermo-measuring sensors are applied to detect the thermal expansion. And a mathematical model is built by multi-variable regression analysis, which is based on the sensed temperature variation and the thermal expansion. Finally, these errors are reduced significantly by sending a feedback to the microprocessor. Meanwhile, the thermal deformation can be compensated by the mathematical model, under the condition when the machine is equipped with linear encoder. Moreover, based on the simulation, it is possible to reduce the number of sensors from ten to six, which save the memory capacity and great benefit for calculating and speeding the process of algorithm. The model improves the accuracy of machining which meets the precision requirement of HSM technology. As the result of various machining tests, the axial positioning error can be reduced from 20 μm to be 3 μm, which is a significantly improvement than existing methods.     

Precision design for machine tool based on error prediction

Digitization precision analysis is an important tool to ensure the design precision of machine tool currently.The correlative research about precision modeling and analysis mainly focuses on the geometry precision and motion precision of machine tool,and the forming motion precision of workpiece surface.For the machine tool with complex forming motion,there is not accurate corresponding relationship between the existing criterion on precision design and the machining precision of workpiece.Therefore,a design scheme on machine tool precision based on error prediction is proposed,which is divided into two-stage digitization precision analysis crucially.The first stage aims at the technology system to complete the precision distribution and inspection from the workpiece to various component parts of technology system and achieve the total output precision of machine tool under the specified machining precision;the second stage aims at the machine tool system to complete the precision distribution and inspection from the output precision of machine tool to the machine tool components.This article serves YK3610 gear hobber as the example to describe the error model of two systems and basic application method,and the practical cutting precision of this machine tool achieves to 5-4-4 grade.The proposed method can provide reliable guidance to the precision design of machine tool with complex forming motion.     

An improved machine tool volumetric error compensation method based on linear and squareness error correction method

Error compensation is one of effective and economical means to improve the machining accuracy of machine tools. The measurement accuracy of kinematic error     

Thermal error modeling of the spindle based on multiple variables for the precision machine tool

Thermal error, especially the one caused by the thermal expansion of spindle in axial direction, seriously impacts the accuracy of the precision machine tool. Thermal error compensation based on the thermal error model with high accuracy and robustness is an effective and economic way to reduce the impact and enhance the accuracy. Generally, thermal error models are built only on temperatures at some points in the spindle system. However, the thermal error is also closely related to other working parameters. Through the theoretical analysis, the simulation, and the experimental testing in this paper, it is found out that thermal error is determined by multiple variables, such as the temperature, the spindle rotation speed, the historical spindle temperature, the historical thermal error, and the time lag between the present and previous times. In order to examine the performance of thermal error models based on multiple variables, two common methods are used for modeling—the multiple regression method and the back propagation network. The data for modeling are collected from experiments conducted on the spindle of a precision machine tool under various working conditions. The modeling results demonstrate that models established based on the multiple variables have better accuracy and robustness. It also turns out that data filtering before modeling can further improve the performance of the models. Therefore, models based on multiple variables with good accuracy and robustness can be very useful for the further thermal error compensation. In addition, by taking relative importance analysis of multiple variables based on standardized regression coefficients, the influence of each variable to the thermal error is revealed. The ranking of coefficients can also be used as a new criterion for the optimal temperature variable selection in the future research.     

基于数字孪生的多轴数控机床轮廓误差抑制方法

针对数控机床高速高精加工过程中轮廓误差控制难的问题,提出一种基于数字孪生的轮廓误差抑制方法,并以数控机床的关键执行部件——多轴进给系统为具体对象,构建面向轮廓误差的"建模-预测-控制"闭环抑制技术框架.该方法针对多轴进给系统多属性交叉耦合的特点,建立其高保真数字孪生体,并通过数字-物理空间的多粒度信息传递,实现跨时间尺度下数字孪生体与物理实体在不同维度的虚实精确同步.通过对数字孪生体进行降阶表征,建立轮廓误差多因素动态影响关系模型,融合多粒度信息实现对轮廓误差的动态预估.基于轮廓误差动态预估结果,提出轮廓误差综合抑制方法,实现对时变运动控制参数下多轴进给系统的插补控制.最后,通过小型三轴数控机床的虚实同步运动实验,验证了所提方法的有效性.     

Simultaneous geometric error identification of rotary axis and tool setting in an ultra-precision 5-axis machine tool using on-machine measurement

This paper presents a method to identify the position independent geometric errors of rotary axis and tool setting simultaneously using on-machine measurement. Reducing geometric errors of an ultra-precision five-axis machine tool is a key to improve machining accuracy. Five-axis machines are more complicated and less rigid than three axis machine tools, which leads to inevitable geometric errors of the rotary axis. Position deviation in the process of installing a tool on the rotary axis magnifies the machining error. Moreover, an ultra-precision machine tool, which is capable of machining part within sub-micrometer accuracy, is relatively more sensitive to the errors than a conventional machine tool. To improve machining performance, the error components must be identified and compensated. While previous approaches have only measured and identified the geometric errors on the rotary axis without considering errors induced in tool setting, this study identifies the geometric errors of the rotary axis and tool setting. The error components are calculated from a geometric error model. The model presents the error components in a function of tool position and angle of the rotary axis. An approach using on-machine measurement is proposed to measure the tool position in the range of 10 s nm. Simulation is conducted to check the sensitivity of the method to noise. The model is validated through experiments. Uncertainty analysis is also presented to validate the confidence of the error identification.     

机床床身结构综合优化设计技术

针对目前常用的内部布置加筋板的箱型床身结构,提出一种同时考虑床身结构的动静态刚度和经济性的综合优化技术。由于床身结构的动静态刚度和质量与支撑垫铁的位置、内部加强筋板布局以及各构件的厚度尺寸有关,提出的综合优化设计技术包括以下3个层次:首先以结构的最大静动态刚度最优为设计目标,对床身结构的垫铁位置进行优化;其次在多工况情况下,以结构的最大静态刚度最优为设计目标,对床身结构的内部加筋板布局进行优化;最后以结构质量最小,以结构的动静态刚度为约束条件,对构件的尺寸进行优化。优化结构表明,采用提出的设计方法,在床身结构质量减小24.48%的情况下,结构的一阶固有频率提高7.83%,结构的导轨静变形有所变差,中间段偏离零线0.5μm左右,但并不影响加工精度,可通过预修正来抵消变形,说明了提出的设计方法的有效性。     

机床大件结构复合优化方法研究

在分析机床广义模块设计和结构智能优化的基础上,论述基于人工神经网络和有限元优化的大件模块结构复合优化方法,给出大件模块结构的尺寸优化设计变量自动搜索寻优计算方法及整机部件间的协调优化方法,实现部件间关键尺寸的快速协调优化,提高产品的性能和设计效率.最后,给出某数控机床的大件结构复合优化案例,说明该方法的有效性.     

考虑电控数据的LSTM神经网络进给轴热误差建模

针对进给轴热误差建模中忽略电控数据和时间序列影响的问题,提出一种考虑温度变化与电控数据的长短期记忆(Long-Short Term Memory,LSTM)神经网络热误差预测模型.以三轴立式加工中心为试验对象,首先对进给轴进行热变形分析,再以温度变化、电控数据为输入样本,建立了LSTM神经网络热误差预测模型,随后通过与仅考虑温度变化的LSTM神经网络,以及同时考虑温度变化与电控数据的BP神经网络进行对比分析,试验论证表明,对数控机床进给轴进行热误差建模时,在考虑温度变化的基础上,进一步考虑电控数据可以提高模型的预测精度和鲁棒性,且在同样输入条件下,LSTM神经网络热误差预测模型相较于BP神经网络有更好的预测精度和鲁棒性.     

考虑电控数据的LSTM神经网络进给轴热误差建模

针对进给轴热误差建模中忽略电控数据和时间序列影响的问题,提出一种考虑温度变化与电控数据的长短期记忆(Long-Short Term Memory,LSTM)神经网络热误差预测模型.以三轴立式加工中心为试验对象,首先对进给轴进行热变形分析,再以温度变化、电控数据为输入样本,建立了LSTM神经网络热误差预测模型,随后通过与仅考虑温度变化的LSTM神经网络,以及同时考虑温度变化与电控数据的BP神经网络进行对比分析,试验论证表明,对数控机床进给轴进行热误差建模时,在考虑温度变化的基础上,进一步考虑电控数据可以提高模型的预测精度和鲁棒性,且在同样输入条件下,LSTM神经网络热误差预测模型相较于BP神经网络有更好的预测精度和鲁棒性.