中文摘要：

针对测量机接触式测头在动态测量过程中精度低这一问题，分析了测头的动态误差来源，并通过标准球的测量实验验证了影响测头动态测量精度的主要因素，其中逼近速率、测杆长度、测端直径是关键的3个影响因素。为了减小测头引起的动态测量误差，引入了RBF神经网络误差补偿模型，从而避免了传统误差模型中复杂的数学关系的推导。在Global Class 9158测量机上对标准球的测量数据建立了训练样本，并对标准环规的测量数据作为测试样本进行误差补偿。测试结果表明经过误差模型补偿修正后测量误差均值从3．5μm减小到1．3μm，并且模型稳定可靠。

英文摘要：

To solve the problem of low precision when touch trigger probe of CMM is in the process of dynamic measurement, the sources of the dynamic error for the probe were analyzed. Through the measurement experiment to the standard ball, the main factors which influence the dynamic error of probe were verified. Among them there are three most important crucial factors： approach speed, stylus length and stylus tip diameter. To reduce the dynamic measurement error caused by the probe, the compensation model of RBF neural network was introduced. It can avoid the derivation of complicated mathematical relationships in the traditional error compensation model. On the Global Class 9158 CMM, the standard ball was measured and the network training data was obtained. The standard ring gauge was measured as the test sample and it was compensated by the error model. The experiment results show that after the compensation by the model the mean measurement error decreases from 3.5 μm to 1.3 μm, and the model is stable and reliable.