AN APPLICATION OF NEURAL NETWORK IN CALIBRATION OF COORDINATE MEASURING MACHINES

Abstract

Two most important requirements of Coordinate Measuring Machines (CMM) are the accuracy and the traceability. However, after a long period of use, errors caused by dynamic forces, thermal expansion, loads, etc can decrease the accuracy as well as the traceability. Therefore, CMM is calibrated to minimize these errors as small as possible. First at all, a geometric error model of CMM is proved mathematically. A method of determining 21 parametric errors by using a Hole Plate then is presented. In addition, a back-propagation algorithm is introduced to approximate parametric errors of all points in the CMM working volume. Finally, the proposed calibration method is demonstrated experimentally.