Abstract: Articulated arm coordinate-measuring machines(AACMM) which can work during a wide temperature range and realize the field measurement are widely used. The improvement on AACMM measurement accuracy leads to more attention towards its thermal deformation error which, meanwhile, must be modified. It briefly analyzes the measurement influenced by thermal deformation error of major AACMM parts from temperature changing. A BP neural network model with a single hidden layer for AACMM thermal deformation error is built. Its inputs are the coordinate values of measuring heads and two measuring-machines temperature parameters and outputs are the difference between the measuring heads coordinate and that at 20 ℃. The data samples are collected from experiments and then are used to exercise and simulate models built. The experimental results verified that to correct AACMM thermal deformation errors based on models built is available and effective.

Key words: Articulated arm coordinate-measuring machines, Error-correcting, Neural network, Thermal deformation