Self learning-empowered thermal error control method of precision machine tools based on digital twin

To improve machining accuracy of complex parts, a self learning-empowered thermal error control method of precision machine tools is presented based on digital twin. The memory of thermal error is theoretically and numerically revealed by error mechanism analysis, and then the applicability of long-short-term memory (LSTM) neural network (NN) in the training of the self-learning error model is proved. To improve the predictive accuracy, the Bayesian optimization algorithm is used to optimize such hyper-parameters as the epoch size, batch size, and the number of hidden nodes of the LSTM NN model. Then the self-learning prediction model of thermal error is proposed based on Bayesian-LSTM NN. The fitting and prediction performance of the proposed Bayesian-LSTM NN is better than that of such models as the LSTM NN with random hyperparameters, back propagation NN, multiple linear regression analysis (MLRA), and least square support vector machine (LSSVM). Finally, the self learning-empowered error control method is proposed based on digital twin, and the Bayesian-LSTM NN error control model is embedded into the self learning-empowered error control framework to realize the real-time thermal error prediction and control. When the predicted thermal error is greater than the preset machining error, the control components are recalculated automatically, and inserted into the machining instructions. It is shown that the machining error can be reduced effectively by the self learning-empowered error control method, which is vital for precision machining of complex parts and improvement of the intelligence level.