基于模糊神经网络的接触式测头动态误差补偿  

Dynamic Error Compensation of Touch Trigger Probe Based on Fuzzy Neural Network

作　　者：郭敬滨[1] 丁航[1] 李醒飞[1] 谭文斌[2] 陈诚[2] 

机构地区：[1]天津大学精密仪器与光电子工程学院,天津300072 [2]天津商业大学机械工程学院,天津300134

出　　处：《纳米技术与精密工程》2016年第4期278-282,共5页Nanotechnology and Precision Engineering

基　　金：精密测试技术及仪器国家重点实验室开放基金资助项目(PIL1407)

摘　　要：测头动态误差严重制约高精度坐标测量机发展,为此,提出基于模糊神经网络的测头动态误差补偿方法以提高测量精度.首先利用三坐标测量机测量标准球和标准环规得到训练样本和测试样本,然后分别使用训练样本和测试样本对接触式测头动态误差进行建模和补偿,最后与BP神经网络模型和静态误差模型进行比较试验.结果表明,经模糊神经网络模型补偿后误差从4.6μm减小至1.3μm,精度提升70%以上;模糊神经网络对测头动态误差具有更好的补偿效果和稳定性.证明模糊神经网络模型能够有效提高测头的动态测量精度.The dynamic error of probe severely restricts the development of high precision coordinate meas-uring machines（CMMs）. For this purpose, a dynamic error compensation method based on fuzzy neural network is proposed to improve the measurement accuracy. Firstly, the training sample and testing sample were obtained by measuring standard ball and standard ring gauge on a CMM. Then the dynamic error of the touch trigger probe was modeled and compensated by using training sample and testing sample, respec- tively. Finally, a comparative test was conducted with the BP neural network model and the static error model. Experimental results show that the error is reduced from 4. 6 μm to 1.3 μm, improving the measuring accuracy by more than 70% after the compensation based on fuzzy neural network model; that fuzzy neural network has better effect and stability for the dynamic error compensation of the probe. The fuzzy neural network model can effectively improve the dynamic measurement accuracy of the probe.

关 键 词：接触式测头 动态误差 人工神经网络 模糊逻辑 补偿 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]