CNC固有误差估计新方法

A NEW METHOD TO ESTIMATE THE INTRINSIC ERROR OF CNC MACHINING

提出了一个估计数控机床固有误差的新方法.与已有利用动态最小二乘(MLS)的方法相比,采用径向基函数(RBF)直接对已有数据进行拟合估计,大大提高了计算效率.还观察到,RBF方法在某个采样半径下误差估计精度总是优于MLS方法,而大于这个采样半径后则MLS方法较好.由此提出了穿越半径的概念与一种基于RBF方法与MLS方法的混合方法,以得到更好的误差估计.实验结果证实了新方法的有效性.

This paper proposes a new method to estimate the intrinsic error of CNC machining. Compared with the existing methods based on moving least square method （MLS）, the new method aims to estimate the error directly using the radial basis function （RBF） and thus boosts the computational efficiency greatly. We observe that the RBF method gives better accuracy prediction when the sampling radius is less than a given value, while the MLS method performs better in other cases. Based on this observation, the concept of critical radius and a hybrid method are proposed to give better error predictions. Experimental results are used to demonstrate the effectiveness of the new method.