Calibration-based thermal error model for articulated arm coordinate measuring machines

The influence of thermal changes in the accuracy of articulated arm coordinate measuring machines (AACMMs) is one of their main error sources. Usually, the effects of this problem are minimized by using low thermal-expansion materials in the arm design or by implementing an empirical error correction model based on the outputs of several temperature sensors placed inside the arm. These models, inherited from temperature correction models for robot arms, are based on the empirical arm positioning error characterization of several temperatures inside its measuring range and by posterior corrections implemented using an average error approximation. Considering this approximation as a correction valid for all the temperature range violates the AACMM calibration conditions, which are defined by a kinematic parameter identification procedure at 20 °C. This implies that, by applying these models, the AACMM will work outside of calibration conditions and will not meet the nominal accuracy values obtained from the identification procedure. In this work a new empirical correction model for thermal errors is presented. This model keeps the calibration conditions established in the parameters identification procedure unaltered. This fact makes it possible to apply other correction models for geometric and non-geometric errors obtained at the calibration temperature after having applied the temperature correction. The algorithm and objective functions used, which are based on a new approach including terms regarding measurement accuracy and repeatability and using the measurements of a ball-beam gauge artifact are shown. The empirical correction model and the arm accuracy and repeatability results before and after correction are also explained, showing an important accuracy improvement in the arm performance for typical arm operation at temperatures different from 20 °C.