基于全息激光干涉测量的零件三维形貌制造误差建模理论与诊断方法

In the research supported by NSFC (50905114), the non-linear dimensional error of product in complex manufacturing processes is taken as the study goal. 1) The Error of Product Quality-Object oriented Petri Net (EPQ-OPN) model has been developed to analyze logically the non-linear dimensional errors. 2) The non-linear explicit mathematical model of dimensional errors has been explored to capture the principle of propagation of dimensional errors. 3) The analysis method for unstructed noise and sample uncertainty has been explored..Eleven papers have been published including eight papers indexed by SCI and three papers indexed by EI, and two pieces of software have been developed and the copyrights have been granted. All of these achievements are supported by NSFC (50905114)..For further research, the 3D surface topography error is taken as study goal and the new hologram laser interferometry metrology device is used. The following achievements will be delivered: 1) The Kriging model of 3D surface topography errors is developed to capture the spatial variation structure and spatial anisotropy, and to model the 3D surface topography errors with spatial dependence. 2) The differential motion vector model of 3D surface topography errors propagation is explored to analyze and predict topography errors the in multistage machining processes. 3) The monitoring method for 3D surface topography errors using non-linear kernel function and mixture Gauss model to on-line visually monitor the process status is developed. 4) Fault diagnosis and process optimization methods are developed using 3D workpiece surface topography texture parameters.

青年基金进展：针对尺寸误差，1)建立尺寸误差面向对象Petri网模型，解决逻辑层上非线性尺寸误差预测问题；2)建立尺寸误差非线性数学模型，揭示非线性尺寸误差关系形成机理；3) 建立了非结构化噪声和抽样不确定性分析方法，提供了控制精度。.青年基金成绩：在青年基金(NSFC: 50905114)资助下已发表学术论文11篇，其中SCI论文8篇，EI论文3篇，取得软件著作权2项。.本项目研究内容：针对形貌误差，采用新的全息激光干涉测量技术:1）建立三维形貌误差克里金建模方法，描述形貌误差空间变异结构和空间各向异性，解决具有高度空间相关性的形貌误差建模问题。2）建立多工序制造过程形貌误差传递微分运动矢量模型，实现多工序过程形貌误差传递分析与预测；3)建立基于非线性核函数和混合高斯模型的形貌误差监控方法，解决形貌误差实时在线可视化监控问题。4)开展基于三维形貌多组织特征参数的故障诊断与工艺优化。

零件表面形貌不但直接影响到零件的润滑性、耐腐蚀性、抗疲劳能力、密封性能等，而且还会对设备的使用寿命、动力消耗、噪声及振动、工作性能等产生较大的影响。因此对机加工表面形貌的准确分析和评定是提高产品质量和使用性能的重要保证。（1）考虑加工因素与零件表面三维坐标之间的空间相关性，建立三维形貌克里金建模方法，描述形貌误差空间变异结构和各向异性，解决了具有高度空间相关性的形貌误差建模问题。实例表明采用单元克里金方法形貌误差估计为真实值的94.21%，考虑加工因素的多元协克里金方法形貌误差估计达到真实值的97%，大大提高了形貌误差估计的分析精度。（2）建立多工序制造过程形貌误差传递模型，实现形貌误差传递分与预测。提出了基于剪切波的零件表面滤波方法，通过非下采样剪切波变换把零件表面将被分解成不同的尺度，每一个尺度上都会得到高频和低频的剪切波系数。而对剪切波系数使用逆非下采样重构表面成分，并根据相应的截止波长将粗糙度、波纹度和形状误差等分离开来。剪切波具有零相移的相位传输特性，因而滤波的结果不会产生相位畸变。此外，剪切波在截止波长附近有比高斯滤波更为陡峭的幅度传输曲面，更接近最理想的幅度传输特性。（3）建立基于非线性核函数和混合高斯模型的形貌误差监控方法。提出了零件表面多类别集成分类方法MPO-SVME，该方法针对缸体顶面的平均分类正确率为92.22%，高于单个分类器的89.58%和集成方法中表现最好的Random Subspace方法90.07%。故提出的MPO-SVME分类方法不仅能取得较高的分类正确率，而且有着较高的运算效率，十分适用于机加工零件表面的识别和检测。（4）提出基于形貌多组织特征参数的故障诊断方法和新的宏观形貌评价方法。. 发表论文23篇，其中SCI论文15篇，EI论文8篇。申请专利4项，授权2项。培养博士生3名、硕士生3名。