【作者】 胡毅；

【导师】 费业泰；

【作者基本信息】 合肥工业大学 ， 测试计量技术及仪器， 2011， 博士

【摘要】 随着生产和科学技术的不断发展,现代制造业正朝着高效率、高精度、高质量及高智能等技术方向发展,对精密测量和精密加工提出了越来越高的要求,精密仪器和精密机床等设备的热变形是制约精密测量和精密加工精度进一步提高的主要因素之一。坐标测量机有着近五十年的发展历史,关节式坐标测量机后期发展而来,但由于其精度高、使用灵活、便携和对使用环境要求低等优点,得到了广泛的推广和应用,在我国的需求也与日俱增。关节式坐标测量机可以工作在10℃~40℃范围内,其热变形引起的测量误差为该机器的主要误差源之一,要实现测量机高精度测量必需对热变形误差加以修正。本文在关节式坐标测量机测量模型基础上,首次对测量机多温度误差源及误差特性进行了研究,并建立了基于神经网络和神经网络集成的关节式坐标测量机热变形误差修正模型,经实验,这对于减小关节式坐标测量机热变形误差有一定的效果。论文在关节式坐标测量机测量模型和误差模型的基础上,对关节式坐标测量机的热特性进行了较为深入的研究,分析了在内、外热源作用下关节式坐标测量机上关节构件、臂、圆光栅传感器等热变形及对测量产生的影响。特别针对关节构件的热变形导致光栅动、定尺相互位置变化,论文采用傅里叶频谱分析方法,建立包含定动光栅位置姿态参数的光场输出模型,给出了定、动栅不平行时输出光场的数学表达式,并在matlab中进行了仿真,给出了不同偏角下的莫尔条纹误差值和及变化趋势。采用模糊聚类分析方法对关节式坐标测量机热变形误差建模及修正中温度测点进行分类,并确定以其中两个测点为参与建模的最佳温度测点。使用神经网络理论对关节式坐标测量机热变形误差数学建模进行了研究分析,以测头三坐标及两温度测点温度为输入、相对于20℃该点的变化量为输出构建了具有单隐层BP神经网络的关节式坐标测量机热变形误差模型。考虑到测量机位姿对测量误差、温度误差的影响,提出了基于神经网络集成的决策级数据融合热变形误差修正模型,该模型中的两个子网络一个是基于空间坐标点的神经网络----在单一姿态下以测量空间点坐标为输入特征变量的热变形误差神经网络模型;另一个是基于测量机位姿的的神经网络----所选择的有限个空间测量点在不同测量位姿下六个角度为输入特征变量的热变形误差神经网络模型。将两网络数据融合以提高模型的泛化能力。成功研制了关节式坐标测量机数据采集系统,该系统包括六个圆光栅传感器的数据采集、两个温度测点上温度传感器的数据采集及与上位机通信的电路,编制了下位机软件。系统采集的数据作为建模的数据样本。建立了关节式坐标测量机温度实验系统。做了关节式坐标测量机上温度场分布实验;通过对一标准杆件的长度测量,给出了关节式坐标测量机在环境温度改变及内热对测量的影响,通过实验表明了必须要对测量机的热误差进行修正;就模型所需数据样本进行了长时间的测量采集,并用此对所建模型进行训练和仿真。实验结果表明所建模型对关节式坐标测量机热变形误差修正是有成效的,特别是融合模型,可使误差达到原误差的二分之一。更多还原

【Abstract】 With the rapid development of manufacturing science and technology, the modern manufacturing industry is developing towards the directions of high- effective, high-precision, high-qualitative and high-intelligent, which brings higher and higher demands on precision measurement and process. The thermal deformation of precision instruments, precision machine tools and other devices is the key restraint for further improving the accuracy of precision measurement and process. Coordinate measuring machines have developed for about fifty years, and articulated arm coordinate measuring machines developed later. However, because of its advantages of high precision, flexible use, low demand for operational environment and portable, articulated arm coordinate measuring machines are being widely used and promoted, also highly demanded in China. These machines can work with temperature ranging from 10℃to 40℃. The measurement error caused by thermal deformation is the major one among all their error sources, so only correcting the thermal deformation error can realize the high precision measurement. Based on the measurement models of articulated arm coordinate measuring machines, multi-temperature error sources and error properties for the measuring machine have been studied first time, and the thermal deformation error correction model based on BP neural network and neural network integration have been built. The experiments show that the work can effectively reduce thermal deformation error of the measuring machine.Based on the measurement models and error models of articulated arm coordinate measuring machines, this paper deeply researched their thermal character, analyzed the influence of measurement errors resulted from the thermal deformation of joint components, arms and circular grating sensors on articulated arm coordinate measuring machine under the action of internal and external heat sources. Especially, for the mutual position change between moving and fixed scales of gratings resulted from joint component deformation, Fourier spectrum analysis method was applied to build the optical field output model involving fixed and moving gratings pose parameters, to introduce the mathematic formulas for output optical field with fixed and moving gratings unparallel mutually. All of which are simulated in Matlab. This paper also gave Moire fringe error values and their variation trend for different offset angle. The fuzzy group analysis method has been adopted to select the temperature measurement points for articulated arm coordinate measuring machines thermal deformation model building and compensation. The two points in which as the best temperature measuring points to join in the modeling have defined.Using neural network mathematical modeling of articulated arm coordinate measuring machines thermal deformation error was studied and analyzed. A BP neural network model with a single hidden layer was built, which inputs were the 3-D measurement data of measuring heads and temperature data at the two temperature measuring points on machine, and outputs were the changes of the space coordinate points compared to that at 20 degree. Considering the influence of measurement errors and temperature errors coming from machines’pose, the paper proposed the thermal deformation error correction model for decision-level data fusion on a basis of neural network integration. This model has two sub-networks, one based on space coordinate points is thermal deformation error neural network model using space point coordinates measured as input characteristic variables under single machine pose, and the other one based on the poses of measuring machines is to use the six angle values of selected limited space measuring points in different measuring poses as input characteristic variables. Fusing the two network data can improve the generalization ability of models.A data collection system for articulated arm coordinate measuring machines was developed successfully, which includes the circuits for data collection of six circle grating sensors and temperature sensors on two temperature measuring points and the circuit for communicating with upper monitor. Software compilation for lower computer was performed. The data collected by the system was used for modeling as the data sample.We also set up the temperature experiment system for articulated arm coordinate measuring machine. The experiment of temperature field distribution of the articulated arm coordinate measuring machine was done. Measuring a standard bar using the measuring machine, its length changed with ambient temperature and internal thermal, that shows the measuring machine thermal deformation error shall be corrected. The data samples needed in models were collected for a long time, and using these samples to exercise and simulate built models. The experimental results certified the effectiveness of correcting the articulated arm coordinate measuring machines’thermal deformation errors by these models set up, especially the fusion models in which the original errors can be reduced by half.更多还原