Reduction of Temperature Measurement Points for an NC Machine Tool Based on Rough Set and Partial Correlation Analysis

To reduce the number of temperature measurement points and improve the efficiency of temperature data acquisition, a new method of temperature measurement based on rough set and partial correlation analysis was proposed. First, based on the way of partial correlation analysis, the partial correlation coefficients between the temperature variables and thermal error of spindle were calculated, and it was used as the basis of choice of the main temperature sensitive variables. Then, the feasible temperature measuring points of the combination by rough sets were obtained, and the most sensitive temperature variables including temperature and partial correlation degree high point combination were screened. Finally, linear regression model of thermal error was established to test prediction accuracy, and verified in a certain type of CNC machine. Results show that temperature sensors are reduced from 22 to 6 to improve the precision and robustness of the thermal error model to a great extent.