The length error compensation method of articulated arm coordinate measuring machine

It is difficult to establish the error model of an articulated arm coordinate measuring machine by using the traditional mathematical modeling because there are many error factors between which the relationship is nonlinear. Taking the angle coding, thermal deformation, probe system, and coordinate values as the research object on the basis of parameter calibration, this paper establishes the compensation model to reduce length error through the Back Propagation (BP) neural network. The connection weights are optimized by the modified simulated annealing (MSA) algorithm so as to solve the issues of neural networks easily falling into a local minimum value and poor anti-interference ability. The data samples obtained through experiments are utilized to train the error model, while the model built compensates test data. It is indicated in the experiment that the average of length measurement after the compensation decreases from -0.0946 mm to -0.0046 mm, and the length measurement deviation at eight different locations is reduced by 0.1352 mm with respect to the MSA-BP model. Published under license by AIP Publishing.