基于 RBF 神经网络的三坐标测量机动态测量误差预测

摘要

When letting CMM run for the X-axis motion, real-time dynamic measurement errors are gathered at four Y-axis positions. Radial basis function (RBF) neural network is applied in modeling and forecasting dynamic measurement errors. The first 100 values of the error sequence of one of the four positions are used to train data and the next 50 values to test the estimation results. The last 50 values of the other three positions are also estimated using the same model. The simulation results show the RBF neural network has a better forecasting accuracy in the single-axis motion. The experiments show that RBF neural network can be applied to estimation of CMMs dynamicmeasurementerrors.%　　三坐标测量机作X单轴运动,测得Y轴4个不同位置的实时动态测量误差,可将径向基函数(RBF)神经网络用于动态测量误差的建模和预测。将其中一个位置采集到的误差序列的前100个值用于RBF神经网络的训练得到模型,后50个值用于预测结果,并用该模型对其他3个位置的后50个测量结果进行预测。仿真结果表明,RBF神经网络对单轴运动的动态测量误差具有较好的预测精度,RBF神经网络可应用于三坐标测量机动态测量误差预测