Identification of rotary axis position errors of CNC machine tools and their analysis using machine learning techniques

Abstract

The paper presents the results of measurements of kinematic centre position errors of rotary axes of five-axis machine tools with a swivel rotary table, using the modern R-Test diagnostic and calibration system. The use of the system was the basis for obtaining experimental data, enabling the assessment of changes in the position of the kinematic centre of the rotary axes of the tested machine tools. The estimation was made on the basis of a statistical analysis of the measurements obtained. The measurements made it possible to determine the position errors of the rotary axes A, B, C in planes Y-Z, X-Z, X-Y and the average values of the calibration ball deviation in specific axes of machine tools. The deviation parameters of the reference ball in the machine tool space were simulated using 3 models - Multi Layer Perceptron, Kolmogorov-Arnold Networks and Multi Output Gaussian Process (MOGP). The simulation results showed that the MOGP model performed best in estimating the deviation parameters. In addition, the MOGP model achieved the best values in all metrics compared to the other models tested. Statistical analysis showed that the largest errors in the position of the kinematic centre of the Lasertec 65 machine tool's rotary axis were recorded for rotary axis A and for the measuring radius R=150 mm (0.099 mm). The smallest errors were recorded for rotary axis C for the same radius R=150 mm (0.0179 mm). The monoBLOCK 65 machine tool demonstrated the smallest error in the position of the kinematic centre of rotary axis A and the minimum measuring radius of R=75 mm (0.0285 mm). The largest error of the monoBLOCK 65 machine tool was recorded for rotary axis B and the measuring radius of R=300 mm (0.0433 mm).The scientific novelty of this work is the development of a complex expert system that enables in-depth analysis of errors in the position of the kinematic centre of the rotating axes of CNC machine tools, as well as forecasting the wear of machine components and detecting anomalies in machine operation. The expert system was developed not only for laboratory users, but also for industrial users for whom maintenance and production quality management are of paramount importance.