SGER: An Integrated Approach to Prediction, Assessment and Inspection of Form Errors in Machined Parts

这个探索性研究（SGER）的目的是表征和相关的加工零件的形状误差与加工工艺序列和工艺变量用于生成的形式。这些信息将用于：（a）提供反馈，并建立明确的指导方针，通过最佳选择工艺顺序和工艺参数来减少形状误差，（B）制定个性化最佳检查计划的方法，以评估每个形状误差的真实值，这些形状误差是工艺顺序/工艺变量特定的，以及（c）制定工艺/机器的形状工艺能力指数。 在本研究中，将考虑两种类型的形状错误，即圆形和不规则性。所考虑的工艺领域将限于由传统的制孔工艺和车削工艺产生的形状特征。这些形状轮廓/表面将使用分析傅立叶模型和广义傅立叶模型进行建模。 通过将轮廓映射到新的s-theta和psi-s域，将提取轮廓/表面的属性，例如一般形状、波瓣数量、波瓣幅度、偏心率、波瓣间距均匀度、形状误差。基于神经网络的方法将用于训练系统，以预测标准化轮廓、标准化属性的范围（波瓣、振幅等）。以及基于工艺顺序和参数的形状误差范围。对于每个标准化轮廓系列，将使用基于实验设计（DOE）的方法来确定样本量、采样方法和形状检查评估算法（最小二乘或最小区域）的最佳组合，该方法针对每个轮廓系列的工艺特定，并且可以估计具有一定置信度的真实形状误差。一种新的组合优化配方结合遗传算法（GA）将被用来计算最小区域的形状误差。 最后，基于标准化的轮廓和每个轮廓族的形状误差，将为每个形状误差开发一个形状过程能力指数。如果成功的话，这项研究将提供一个基础上预测的过程变量和序列的基础上形成的轮廓和错误。这可能导致选择最佳工艺计划/变量以减少形状误差。 这项研究还将为评估特定于创建表单的过程的表单错误的最佳检查计划提供指导方针。最后，形成的过程能力指数可用于优化制造公差分配，设计中心和机器选择。这样的计划可以很容易地被设计师和车间的从业人员使用。

The objective of this Small Grant for Exploratory Research (SGER) is to characterize and correlate form errors in machined parts with machining process sequences and process variables used for generating the form. This information will be used to: (a) Provide feedback and establish clear guidelines for reducing form errors by optimal selection of process sequences and process parameters, (b) Develop methods for individualized optimal inspection plans for assessing the true value of each form error that are process sequence/ process variables specific, and (c) Develop form process capability indices for processes/machines. For this research, two types of form errors, namely, Circularity and Cylindricity, will be considered. The domain of processes considered will be limited to form features generated by conventional hole making processes and turning processes. These form profiles/surfaces will be modeled using analytical Fourier Models and Generalized Cylinders. Attributes of profiles/surfaces such as general shape, number of lobes, amplitude of lobes, eccentricity, evenness of lobe spacing, form errors, will be extracted by mapping profiles to a novel s-theta and psi-s domains. A Neural network based approach will be used for training the system to predict standardized profiles, ranges of standardized attributes (lobes, amplitude etc.) as well as range of form errors based on process sequence and parameters. For each family of standardized profiles, a Design of Experiments (DOE) based approach will be used for determining the optimum combination of sample size, sampling method, and evaluation algorithm (least square or minimum zone) for form inspection that is process specific for each profile family, and can estimate the true form error with a certain level of confidence. A novel combinatorial optimization formulation in conjunction with Genetic Algorithms (GA) will be used to calculate minimum zones of the form errors. Finally, based on standardized profiles and form errors for each profile family, a form process capability index will be developed for each of the form errors. If successful, this research will provide a basis for predicting form profiles and errors based on process variables and sequence. This could result in selection of optimal process plans/variables for reducing form errors. This research will also provide guidelines for optimal inspection plans for assessing form errors that are specific to the process creating the form. Finally, the form process capability indices developed can be used for optimal manufacturing tolerance allocation, design centering and machine selection. Such a plan can readily be used by designers as well as practitioners on the shop floor.