Mechnical Design Systems Laboratory, Hiroshima University

Automated and robotized manufacturing systems have been required as a countermeasure for lack of labor force in manufacturing industry. Machining accuracy is the most essential factor for continuous machining operation. Thermal deformation due to machine’s internal heat generation or heat exchange with the ambient environment deteriorates machining accuracy. To suppress the thermal displacement, various methods have been proposed. However, these methods fail to compensate machines in severe situations such as unexpected temperature change and sensor failure. In this paper, an adaptive thermal displacement compensation function for turning center using deep learning is proposed. The function predicts thermal displacement from the temperature of the machine structure. The function is applicable to the conditions where machine’s environment temperature changes or heat is generated during machining operation. In addition, prediction uncertainty is calculated based on deep learning to adaptively change the compensation weight. This algorithm suppresses deterioration of accuracy in the event of large prediction uncertainty. High-performance thermal displacement compensation result on a turning center is presented.

Keywords: Precision Machine, Thermal error, Compensation, Deep learning