Neural Network Error Compensation of Machine Tools and Coordinate Measuring Machines

机床和三坐标测量机的神经网络误差补偿

• 批准号: 9017293
• 负责人: John Ziegert
• 金额: --
• 依托单位: University of Florida
• 依托单位国家: 美国
• 项目类别: Continuing grant
• 财政年份: 1991
• 资助国家: 美国
• 起止时间: 1991-02-15 至 1995-01-31
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=9017293&HistoricalAwards=false
• 关键词: Neural; Network; Error; Compensation; Machine

This award supports research involving a study of the applicability of Artificial Neural Networks to the problem of determination and compensation of positioning errors in machine tools and coordinate measuring machines. Research issues which will be explored are: (1) Prediction and compensation of geometrically induced errors using a neural network to map workspace positions into tool point error vectors with training data obtained from machine measurement of a precision physical artifact. (2) Prediction and compensation of thermally induced errors using a small number of strategically located temperature probes as additional input to an expanded neural net, and performing the artifact measurement for differing thermal states of the machine. (3) Generation of design rules for the design of the physical artifact to insure optimal performance of the neural net error compensation technique. (4) Investigation of the possibility of using the results of dimensional inspection of manufactured parts to update the training set for the neural network; thus closing the loop between process output quality and process input variables, resulting in continuous process improvement in an automated factory environment. (5) Investigation into the necessary architectural features and computational capabilities of a hybrid neural net machine tool controller.

该奖项支持涉及人工神经网络在机床和坐标测量机定位误差的确定和补偿问题中的适用性的研究。 将探讨的研究问题是：（1）使用神经网络预测和补偿几何引起的误差，将工作空间位置映射到工具点误差向量，并使用从精密物理工件的机器测量获得的训练数据。 (2) 使用少量策略性定位的温度探针作为扩展神经网络的附加输入来预测和补偿热引起的误差，并对机器的不同热状态执行伪影测量。 (3)生成用于物理工件设计的设计规则，以确保神经网络误差补偿技术的最佳性能。 (4) 研究使用制造零件的尺寸检验结果来更新神经网络训练集的可能性；从而形成过程输出质量和过程输入变量之间的闭环，从而在自动化工厂环境中实现持续的过程改进。 (5) 研究混合神经网络机床控制器的必要架构特征和计算能力。