Thermal error compensation method for CNC machine tools based on deep convolution neural network

Abstract To avoid the resetting of temperature-sensitive points caused by seasonal and weather changes influence on accuracy thermal error model, a prediction compensation method for CNC machine tools without preselected was proposed based deep learning method. The original temperature data tool were converted into images directly serve as input network. improve model with 10 hidden layers constructed using powerful image feature ability convolution nonlinear mapping relationship between pre-selection key established, more relationships retained. STM32F4 microprocessor used controller to develop software hardware system. application experiment system is carried out grinder, effectiveness verified.